arXiv:2607.15105v2 Announce Type: replace
Abstract: Parameter-efficient fine-tuning reduces model and optimizer memory, but dense attention still makes long training sequences expensive. We combine Hierarchical Global Attention (HGA) with segment-wise backpropagation and tiered KV storage. Only the active segment remains differentiable in VRAM; older KV is detached into RAM or NVMe, and HGA loads a bounded set of exact historical tokens for each query block. On Qwen3-8B with 4-bit QLoRA and PG19, dense training on a 16 GB Quadro RTX 5000 fits 2,048 tokens but fails at 4,096, whereas HGA reaches 16,384 tokens with 15.28 GB peak VRAM. Under evaluation the same adapter runs through 131,072 tokens on this card; VRAM is not constant but grows gently with the resident chunk summaries, so RAM and NVMe capacity set the practical limit beyond these lengths. At the shared 2K training length, HGA-trained and dense-trained adapters obtain 2.7405 and 2.7383 nat under the same dense-attention readout, while the stock model obtains 2.9541. At this boundary HGA training is already marginally faster (217.75 vs. 207.02 tokens/s), and the HGA-to-dense throughput ratio improves from 1K to 2K; because HGA keeps the attended historical set per token approximately constant while dense work per token grows, we expect this lead to widen as context grows. Dense attention is used for the main quality and retrieval comparisons so that they measure the learned weights and remain compatible with standard generation frameworks. HGA can also be used for retrieval and generation; an optimized production-grade serving implementation is under development.
Science Journals
arXiv:2607.15178v2 Announce Type: replace
Abstract: Transformer reasoning is limited by autoregressive decoding, which repeat edly compresses rich hidden computation through token space and makes it difficult for intermediate reasoning states to persist across time. We in troduce Transformers with Temporal Middle-Layer Recurrence (T2MLR), a transformers-based latent reasoning architecture that fuses a cached middle layer representation from the previous token directly into an earlier layer of the current token position, enabling abstract intermediate computation to persist across decoding steps with little inference overhead. Across natural-language pretraining and multi-hop reasoning finetuning, T2MLR consistently outperforms data- and parameter-matched Transformer base lines. Moreover, applying recurrence to only a localized middle-layer block (as little as 20% of the network) often outperforms full-layer recurrence. Im portantly, T2MLR does not require pretraining from scratch: retrofitting the recurrent pathway into an existing pretrained 1.7B Transformer and briefly finetuning substantially improves math reasoning, lowering the barrier to practical adoption. These results suggest that effective latent reasoning in Transformers does not require looping over all layers as in previous works, but can instead emerge more strongly from targeted middle-layer recurrence.
arXiv:2607.15190v2 Announce Type: replace
Abstract: AI benchmarks increasingly leverage item-level statistical models, particularly item response theory (IRT), to estimate model capabilities, rank systems, select informative examples, and diagnose benchmark quality. However, AI benchmark data often departs from the data regime of human testing, for which standard IRT estimation tools were originally developed: benchmarks typically involve fewer evaluated models, far more items, and capability distributions that may be skewed, clustered, or multimodal. We examine how these regime mismatches challenge the reliability of IRT modeling for AI evaluation. Using item parameters and capability distributions derived from six widely used LLM benchmarks, we simulate response matrices under three common IRT models and compare four estimation tools used in recent benchmark studies: marginal maximum likelihood, Markov chain Monte Carlo, variational inference, and a neural pseudo-Siamese estimator. Across 18,000 simulation conditions, we systematically evaluate computational feasibility, scalability, and the reliability of IRT inferences about model rankings, predicted performance, and item characteristics. Results show that classical estimators can become infeasible in large benchmark settings, whereas scalable estimators can produce unreliable item-level and ranking inferences with small or non-normally distributed model sets. This study identifies when latent trait models reliably support or risk distorting AI benchmarking claims, and what sample sizes and diagnostics are needed for trustworthy use.
arXiv:2607.15263v2 Announce Type: replace
Abstract: Security-agent evaluations commonly measure peak offensive capability under generous inference budgets, emphasizing vulnerability discovery, exploit development, penetration testing, and CTF completion. Such measurements are useful but incomplete: in operational security, every reasoning step, tool call, telemetry query, and enrichment request consumes budget. We evaluate language-model security agents through this cost-success lens on offensive Cybench challenges and defensive Splunk BOTS v1 investigation challenges. Instead of reporting only best-case success, we compare models at fixed cost levels and decompose performance by inference spend and tool spend. Our results show distinct scalingregimes for red- and blue-team tasks. Offensive CTF performance improves with additional test-time compute, and scaled open-weight models can approach frontier proprietary systems while remaining cost-competitive. Defensive SOC investigation does not scale in the same way: success depends more heavily on disciplined tool use, telemetry navigation, and selective enrichment than on raw reasoning budget alone. We argue that security-agent benchmarks should measure economic efficiency and operational fit alongside task success. Cost-aware, SOC-native evaluations provide a clearer picture of which models are practically useful today and where defensive agents still need to improve. We present an interactive website with our results https://evals.frontier.security.
arXiv:2402.17500v2 Announce Type: replace-cross
Abstract: A central question of network science is how functional properties of systems emerge from their structure. For networked dynamical systems, structure is typically captured through network measures. We investigate the relationship between these measures and stability metrics across non-linear and linear oscillators, as well as real-world power grid topologies and dynamics. We find that this relationship is highly sensitive to the underlying ensemble: minor changes in the networks considered, such as going from mean degree 6 to mean degree 8, can invert the correlation between a network measure and stability. We also investigate network measures as inputs for machine learning, as well as Graph Neural Networks (GNNs) as predictors of stability. Both GNNs and the non-linear combination of many network measures can accurately predict stability within a given ensemble, yet both can fail when the ensemble changes. We conclude that neither approach reliably identifies the underlying structural causes of instability.
arXiv:2506.24007v5 Announce Type: replace-cross
Abstract: This study investigates minimax and Bayes optimal strategies for fixed-budget best-arm identification. We consider an adaptive procedure consisting of a sampling phase followed by a recommendation phase, and we design an adaptive experiment within this framework to efficiently identify the best arm, defined as the one with the highest expected outcome. In our proposed strategy, the sampling phase consists of two stages. The first stage is a pilot phase, in which we allocate samples uniformly across arms to eliminate clearly suboptimal arms and to estimate outcome variances. Before entering the second stage, we solve a Gaussian minimax game, which yields a sampling policy and a decision rule. In the second stage, samples are allocated according to this policy. After the sampling phase, the procedure enters the recommendation phase, where we select an arm using the decision rule. We prove that this single strategy is simultaneously asymptotically minimax and Bayes optimal for the simple regret, and we establish upper bounds that coincide exactly with our lower bounds, including the constant terms. The lower bounds hold against every adaptive experiment and for every fixed number of arms, and the strategy attains them without knowing the outcome distributions or the prior.
arXiv:2512.15847v2 Announce Type: replace-cross
Abstract: The observation of filamentary X-ray structures near bow-shock pulsar wind nebulae (PWNe) -- such as the Guitar, Lighthouse, and PSR J2030$+$4415 nebulae -- and of slow-diffusion regions around pulsars like Geminga, Monogem, and PSR J0622$+$3749, challenges the standard picture of cosmic-ray transport in the interstellar medium, implying a diffusion coefficient two orders of magnitude smaller than the Galactic average. The suppressed diffusion can be attributed to self-generated magnetic turbulence, driven -- via the non-resonant streaming instability -- by electron-positron pairs escaping the PWNe. This instability requires a net current, yet the beam of escaping pairs is expected to be charge-neutral. We show that a charge-neutral pair beam propagating through an electron-proton plasma can spontaneously generate a net current. Using fully kinetic two-dimensional particle-in-cell simulations with realistic mass ratio, we find that beam electrons get focused into self-generated magnetic filaments produced by the nonlinear evolution of the Weibel instability, while beam positrons remain unconfined. We show that in three-dimensional simulations the resulting net (positron) current drives the non-resonant streaming instability, further amplifying the magnetic field. This mechanism provides a pathway for the onset of charge asymmetries in initially charge-neutral pair beams and for the growth of magnetic fluctuations that efficiently scatter the beam particles, with implications for the formation of X-ray filaments and, potentially, for particle self-confinement in TeV halos around PWNe.
arXiv:2604.16183v2 Announce Type: replace-cross
Abstract: Reliable prediction of the solar cycle is a formidable challenge, yet it is increasingly vital in our technology-dependent society as solar activity drives space weather. Various methods, including precursors, nonlinear curve fitting and extrapolation, statistical and Machine Learning (ML) models, and dynamo and surface flux transport (SFT) models, were implemented to predict past cycles. Analysing about 100 predictions for Solar Cycle 24 and over 130 for Solar Cycle 25, we find that most methods largely failed to predict the peak correctly: Cycle 24 was statistically predicted to be a strong cycle, whereas Cycle 25 was predicted to be a weak cycle. By and large, predictions made only after the cycle began became closer to reality. ML-based models also produced discouraging results. The polar field and its proxy-based predictions are the most physically supported approach to prediction; however, applying them much earlier, before the solar minimum, may yield inaccurate results. Dynamo models are progressively improving both in understanding and in forecasting; however, they need to improve by accurately assimilating the observed polar field data and additional physics, such as meridional flow variations. Solar dynamo theory, complemented by the SFT model and observations, demonstrates that the prediction of a cycle before the time of its previous cycle's maximum is meaningless. The current solar cycle is declining, and the community is now preparing for the prediction of the next cycle. Thus, this review will guide future studies.
arXiv:2604.21274v3 Announce Type: replace-cross
Abstract: A random access code (RAC) encodes an $L$-bit string into a $k$-bit message, $L>k$, so that any requested bit can be recovered with high probability; a quantum RAC (QRAC) uses $k$ qubits instead. We give a geometric characterization of optimal classical $(L,k)$-RACs under average and worst-case decoding criteria. The average criterion is reduced to choosing $2^k$ representatives in $\{0,1\}^L$, while the worst-case criterion is reduced to a minimax problem over $2^k$ points in $[0,1]^L$ with a distance-like objective. This framework proves optimality for several parameter families, with many optimal constructions arising from standard infinite families of binary linear codes. It also yields two explicit classical--quantum separations. First, for every $L>1$, we construct a $(L,1)$-QRAC whose average decoding success probability strictly exceeds the optimal classical value. Second, for the family $(2^k-1,k)$, we prove worst-case optimality of a classical RAC and construct a QRAC with strictly larger worst-case success probability. For the family $(L,L-1)$, the framework identifies a classical RAC that is average-case optimal and, under a stated conjecture, also worst-case optimal. The same viewpoint further recovers explicit $(L,L-1)$-QRACs attaining a previously conjectured upper-bound value.
arXiv:2605.13466v3 Announce Type: replace-cross
Abstract: We present the results of an experimental study of the anomalous anisotropy of alignment signals in cesium vapors under strong spin-exchange conditions near zero magnetic field with linearly polarized optical pumping. We show that the anisotropy of the Hanle resonances in the plane perpendicular to the pump beam increases with concentration: in one direction the widths remain broadened by spin-exchange, whereas in the other they approach the spin-exchange relaxation free limit. With a further increase in concentration, additional nonlinear effects arise, such as signal amplification, bistability, hysteresis, and memory. To explain these effects we construct a illustrative theoretical model incorporating spontaneous polarization effects under strong spin exchange conditions. The model qualitatively shows that the ultra-narrow alignment resonances may originate from quadrupole anisotropy arising from the projection of spontaneous transverse orientation onto the detection axis. The unique properties of these resonances, such as their extremely small width and magnetic field-controlled bistability with a long-term memory effect, make them promising for use in quantum sensing and information.
arXiv:2605.16560v3 Announce Type: replace-cross
Abstract: In this work we consider a dynamical cellular communication network in which mobile base stations (BSs) are modeled as a homogeneous Poisson point process on $\mathbb{R}^2$. Each base station moves at a constant speed in a random direction. A typical user connects to the nearest base station and it experiences variable signal and interference powers depending on the distance of all the stations. Along the motion of the stations, the user swaps its serving station, and such an event is called a {\em handover}. We are interested in the performance evaluation of the system under some classical and tropical metrics of interest at different time of events, inducing handovers, maximal proximity of serving station, nearest interferer at closest or farthest distance with respect to the user or at any typical time epoch. The main results of the paper are closed or integral form expressions for the basic metrics of interest, in particular coverage probability and Shannon rate at these epochs. We can make an analogy with ``seasons'' based on the fluctuations of signal and interference power. Strong or mild signal or interference power correspond to different seasons of Shannon rate along the evolution of the system. We also provide a complete comparison study of the metrics at interest at these epochs.
arXiv:2607.15505v1 Announce Type: cross
Abstract: Fractional gradient descent (FGD) incorporates long-range memory through Caputo-type operators and has been shown to improve stability in ill-conditioned and nonconvex optimization problems. Despite these advantages, its practical use remains limited, mainly due to the high computational cost of evaluating history-dependent convolutions, which scales quadratically with the number of iterations. In this paper, we focus on making Caputo-based optimization computationally viable without sacrificing its intrinsic memory structure. We begin by expressing the fractional descent direction as a discrete convolution over past gradients, which provides a unified view of the method. Based on this formulation, we introduce two complementary mechanisms to reduce the cost of the memory term. The first uses a sum-of-exponentials (SOE) approximation of the power-law kernel, leading to efficient recursive updates. The second approach, newly proposed in this paper as dyadic hierarchical discrete convolution (DHDC), compresses the gradient history through a multiscale aggregation strategy. Rather than treating these approximations as purely numerical accelerations, we interpret them as perturbations of the ideal Caputo operator. This viewpoint allows us to analyze how the compressed memory affects the optimization dynamics. Under standard $\mu$-strong convexity and $L$-smoothness assumptions, we show that the resulting method still exhibits monotone descent and linear convergence, provided that the approximation error remains controlled.
arXiv:2607.15547v1 Announce Type: cross
Abstract: Conventional pinching antenna (PA)-assisted integrated sensing and communication (ISAC) architectures typically assume static receiver locations or predetermined receive waveguides, thereby underutilizing the inherent spatial degrees of freedom. This paper proposes a novel mode-selectable PA-assisted ISAC framework to maximize the post-combining sensing signal-to-noise ratio while satisfying multi-user quality-of-service constraints by jointly optimizing the waveguide mode selection, transmit beamforming, and transmit/receive PA positions. To tackle the resulting mixed-integer nonconvex optimization problem, we develop a low-complexity block-coordinate descent algorithm that leverages a penalty-based majorization-minimization method to achieve high-quality suboptimal solutions. Numerical results demonstrate that the proposed design significantly outperforms both traditional PA and fixed-antenna benchmarks by synergistically harnessing spatial adaptability and modal reconfigurability. In particular, the mode-selectable design enables the coordinated optimization of transmit/receive operations and sensing-communication resource allocation, thereby maintaining sensing robustness under stringent communication requirements.
arXiv:2607.15575v1 Announce Type: cross
Abstract: We propose an integrated sensing and communications (ISAC) framework that supports chirp signal transmission in CP-OFDM-based multiple access communication systems, enabling efficient coexistence of communication and sensing capabilities. Our framework employs the discrete Fourier transform phase rotated and permuted frequency division multiple access (DFT-p-FDMA) waveform to transmit chirp signals using a portion of the frequency resources, while ensuring interference-free concurrent CP-OFDM data transmissions on other bands. We analyze the effective channel behavior under the DFT-p-FDMA waveform, characterizing how delays and Doppler shifts impact radar target echoes. We also show how processing multiple received symbols improves Doppler resolution in practical scenarios. Our framework allows flexible adjustment of range-Doppler resolution through optimized time-frequency resource allocation, offering a versatile solution for ISAC applications. Simulation results validate the framework's performance in delay and Doppler estimation, highlighting its potential to support ISAC in next-generation wireless networks.
arXiv:2507.21018v2 Announce Type: replace
Abstract: Automated assessment of human motion plays a vital role in rehabilitation, enabling objective evaluation of patient performance and progress. Unlike general human activity recognition, rehabilitation motion assessment focuses on analyzing the quality of movement within the same action class, requiring the detection of subtle deviations from ideal motion. Recent advances in deep learning and video-based skeleton extraction have opened new possibilities for accessible, scalable motion assessment using affordable devices such as smartphones or webcams. However, the field lacks standardized benchmarks, consistent evaluation protocols, and reproducible methodologies, limiting progress and comparability across studies. In this work, we address these gaps by (i) aggregating existing rehabilitation datasets into a unified archive called Rehab-Pile, (ii) proposing a general benchmarking framework for evaluating deep learning methods in this domain, and (iii) conducting extensive benchmarking of multiple architectures across classification and regression tasks. All datasets and implementations are released to the community to support transparency and reproducibility. This paper aims to establish a solid foundation for future research in automated rehabilitation assessment and foster the development of reliable, accessible, and personalized rehabilitation solutions. The datasets, source-code and results of this article are all publicly available.
arXiv:2605.25878v2 Announce Type: replace-cross
Abstract: Pathological assessment guides lung cancer diagnosis, treatment selection, and prognostic evaluation, yet current CPath approaches rely on task-specific models for isolated objectives. Although pan-cancer foundation models offer versatility, they lack subspecialty-level depth and have not been evaluated across clinical workflows or prospectively validated in real-world settings. We introduce PulmoFoundation, a multi-center, prospectively validated, randomized controlled trial (RCT)-evaluated foundation model for comprehensive lung pathology assessment across pre-operative, intra-operative, and post-operative care. Built upon Virchow2 via subspecialty-specific pretraining using ~40,000 diagnostic H&E-stained whole-slide images (WSIs), PulmoFoundation was systematically evaluated on ~26,000 WSIs across 32 clinically relevant tasks. In addition to accurately predicting molecular markers and patient survival, our model achieves clinical-grade performance in core diagnostic tasks across biopsy, frozen section, and surgical resection slides. In a registered prospective study of 1,357 patients across 11 diagnostic tasks, our model achieved an average AUC of 92.3%. Using pre-specified triage thresholds, PulmoFoundation could reduce additional second-review burden for 68.8% of biopsies and 83.0% of frozen sections, and defer 44.5% of IHC stain orders, with PPVs of 1.000, 0.991, and 0.966. Beyond prospective validation, we conducted a crossover RCT with eight pathologists, in which AI assistance improved diagnostic accuracy across 5,264 case-reader pairs (91.7% w/ AI vs. 83.2% w/o AI). AI assistance also reduced median diagnostic time by 18.3%, increased diagnostic confidence by 9.0%, and improved inter-rater agreement from moderate (kappa = 0.55) to substantial (kappa = 0.76). Together, these evaluations support PulmoFoundation as a clinically validated decision-support system for lung pathology.
arXiv:2606.05953v3 Announce Type: replace-cross
Abstract: Tennant claims that his Core logic $\mathbb{C}$ is paraconsistent. It means that the sequent of the First Lewis Paradox, i.e. $\lnot A, A \vdash B$ is declared false, and its corresponding antisequent, called `Claim~1', i.e. $\lnot A, A \nvdash B$ true, as in minimal logic $\mathbf{M}$. This paper proves that Claim~1 entails a contradiction in $\mathbb{C}$, so that, to preserve consistency, the Core logician must reject the claim that his system is paraconsistent. The proof is purely logical, in four steps within a five-rule fragment $\mathcal{F}$ of $\mathbb{C}$ and its refutation system in the sense of Lukasiewicz and Goranko; the Appendix certifies every step in Coq -- with no axiom assumed and every commitment displayed as a named hypothesis -- and the same certification is replayed independently in Lean~4.
arXiv:2606.14440v2 Announce Type: replace-cross
Abstract: Solar energetic particle (SEP) events, defined by proton flux exceeding 10 pfu in the > 10 MeV channel, pose major risks to spacecraft operations, astronaut safety, and high-latitude aviation. Due to the complexity and rarity of SEP events, reliable operational SEP forecasting remains an important challenge in space weather. Here we present a novel 24-hour-ahead realtime forecasting framework, SEPNET-PRISM, based on a multi-task learning structure and a thoroughly constructed list of features from multiple sources spanning multiple solar cycles, that jointly predicts SEP event occurrence and future proton and soft X-ray fluxes. SEPNET-PRISM extends the earlier-introduced SEPNET-based models by integrating a broader range of solar observations, including active-region magnetic parameters from SHARP and SMARP, solar-flare information, coronal mass ejections, soft X-ray flux, and historical > 10 MeV proton flux. As compared with SEPNET, the inclusion of SMARP data expands the temporal coverage of magnetic-field predictors to earlier solar cycles, while flux-based inputs provide additional precursor information. Evaluation on the CLEAR SEP benchmark dataset shows improved classification performance over the earlier SEPNET-O (operational version of SEPNET) on the newly aligned dataset. The best operational model is obtained when magnetic, radiative, and proton-flux predictors are combined, highlighting the value of expanded historical coverage and complementary precursor information for improving realtime SEP forecasting.
arXiv:2606.19585v3 Announce Type: replace-cross
Abstract: Electric field noise produced by the surface of ion trap electrodes reduces the fidelity of quantum computing operations. Despite decades of investigation its microscopic origins remain unclear. Here, we measure electric field noise at trapping locations along the symmetry axis of a linear surface Paul trap. We find that noise levels vary by three orders-of-magnitude in one 600$\,\mu$m section of the trap. Optical and scanning electron microscope images show micron-sized particles close to the trapping locations with the highest noise levels. We find that modeling the particles as a lossy dielectric with a effective loss tangent $\tan\theta=0.33(0.06)$ describes the magnitude of the noise, as well as its spatial and frequency dependence. Our observations may explain the large variation of reported noise levels in literature.
arXiv:2606.20178v3 Announce Type: replace-cross
Abstract: We use density functional theory and model Hamiltonians to reveal large spin splitting of bands localized at ferromagnetic surfaces of bulk antiferromagnets (AFMs). There is great interest in material platforms combining the robustness and ultrafast dynamics of AFMs with large, functional spin splitting which is often restricted to ferromagnets (FMs). Here, we show that a subset of AFM \textit{surfaces} which have symmetry-allowed magnetization can host large spin splitting via bulk degeneracy lifting of sublattice-resolved exchange splittings. We find that the spin splitting is maximized for two ferromagnetic surface motifs: terminations with single uncompensated magnetic sublattices, and two-sublattice surfaces whose sublattices are magnetically compensated in the bulk, but acquire distinct crystal field environments via surface truncation. The latter case can yield FM-like spin splitting magnitudes while also having small uncompensated magnetization. We confirm these predictions with first-principles calculations of $\mathrm{Cr_2O_3}$ and $\mathrm{FeF_2}$, finding splittings as large as $1~\mathrm{eV}$ depending on the surface in question. Our findings point to intrinsic surface symmetry breaking as a route to large, functional spin splitting in an expanded range of AFM materials.
arXiv:2607.00014v2 Announce Type: replace-cross
Abstract: Since times immemorial, total solar eclipses have inspired awe and wonder. In the modern scientific era, they have transformed into exclusive natural laboratories, offering fleeting but invaluable opportunities to study the Sun's faint outer atmosphere that is otherwise obscured by the intense glare of the photosphere. This unique vantage point has enabled revolutionary discoveries, from the identification of the element Helium and the first empirical validation of Einstein's General Relativity, to deciphering the corona's surprisingly high temperature. Today, ground-based eclipse experiments provide crucial data that complements and calibrates our space-based solar observatories, and offer high-resolution capabilities in the spatial, temporal as well as spectral domains. This chapter serves as a comprehensive guide detailing how to leverage modern observing equipments, detectors, and advanced computational techniques in image and data processing to conduct meaningful scientific investigations, bridging the gap between historical precedent and cutting-edge research.
arXiv:2607.03153v2 Announce Type: replace-cross
Abstract: As particle physics detectors grow in scale, High Energy Physics experiments must process ever-increasing data volumes. Level-1 trigger systems, implemented on Field-Programmable Gate Arrays and increasingly using neural-network algorithms, filter this data in real time. However, their proximity to the interaction point exposes them to radiation, which can corrupt outputs, stall processing pipelines, or damage hardware, with significant financial and scientific consequences. In this work, we present the first Register Transfer Level fault-injection study of a deployed Level-1 hardware neural-network trigger, GNN-ETM in the Belle II trigger system. We target three failure modes most consequential to a real-time trigger pipeline: deadlocks, timeouts, and packet-integrity violations. Through two complementary campaigns, we inject 1 442 840 Single-Event Upsets across 211 245 signals. We find a monitoring asymmetry in the existing verification infrastructure and propose inter-stage liveness monitoring as a more accurate alternative to output-only observation, showing that Mean Time To Failure estimates from the two approaches differ by up to 78.7%. The resulting per-stage data identifies the highest-priority hardening targets.
arXiv:2607.15681v1 Announce Type: new
Abstract: Microservice applications are increasingly deployed across cloud--edge environments, where heterogeneous nodes and time-varying inter-node delays amplify the impact of placement decisions. At the same time, these applications face non-stationary traffic, shifts in the mix of root request operations that exercise different call graphs, and heterogeneous communication modes that determine how network latency and queuing propagate to end-to-end (E2E) performance. Existing autoscalers and network-aware schedulers typically handle only a subset of these dynamics, leading to either compute bottlenecks or inflated cross-node latency and thus SLO violations.
We propose ADASCALE, an adaptive framework that jointly scales and places microservice replicas under such multi-dimensional dynamics. ADASCALE implements a Monitor--Analyzer--Planner--Executor (MAPE) loop that extracts per-edge and per-service demand from distributed traces and service-mesh metrics, identifies the most critical root operation under a mixed workload, computes SLO-aware replica targets, and then places replicas to minimize a demand-weighted latency objective given the current inter-node latency matrix. To react quickly to networking perturbations, ADASCALE triggers a reactive placement loop, while a steady-state autoscaling loop handles demand shifts.
We evaluate ADASCALE on a cloud--edge Kubernetes cluster using the DeathStarBench Social Network application with three root operations under varying load and workload mixes. Across scenarios, ADASCALE consistently meets SLO targets and improves both latency and throughput: compared with NetMARKS_Scale, it achieves up to 1.56x, 1.93x, and 1.34x lower average response time (for compose-post, read-home-timeline, and read-user-timeline) and up to 2.16x, 1.32x, and 1.36x higher throughput, respectively.
arXiv:2607.15582v1 Announce Type: new
Abstract: Soft continuum robots require embedded sensing for proprioception and contact detection, yet integrating sensors into sparse, highly deformable architected structures remains challenging. We present a model-based strategy that decouples proprioceptive and contact signals from a common set of fluidic pressure sensors embedded in a soft architected segment. Each segment of the Innervated Trimmed Helicoid (ITH) contains six air channels routed in a localized zigzag pattern along the circumference. With only three principal kinematic degrees of freedom (axial compression, bending in x, bending in y), the six pressure readings form an overdetermined system. A piecewise constant curvature model maps pressures to shape, and Huber regression identifies outlier channels whose residuals indicate external contact. On a single ITH segment, this approach achieves proprioceptive shape estimation with a relative bending error of 0.11 +/- 0.02 and a contact detection rate of 97% across 178 trials. We integrate eight ITH segments into Air-Helix, a tendon-driven soft continuum manipulator, and present exploratory whole-arm demonstrations that include tactile teaching by demonstration, admittance-controlled force regulation, and tactile object reconstruction. The results suggest that localized fluidic innervation combined with model-based redundancy resolution is a practical path toward concurrent proprioception and contact sensing in architected soft robots.
arXiv:2607.15682v1 Announce Type: new
Abstract: Sampling from an unnormalized Boltzmann density requires proposals that move probability mass globally while retaining enough path-probability information for statistical correction. We introduce Neural Non-Equilibrium Hamiltonian Monte Carlo (NHMC), a train-then-correct learned Hamiltonian sampler. Starting from a tractable base distribution, NHMC learns stochastic Hamiltonian-style paths toward the target. Once training is complete, the learned proposal parameters are fixed; the proposal then generates complete paths and endpoint configurations, which are statistically corrected using the recorded non-equilibrium work. This dimensionless generalized work is determined by the probability ratio between the forward proposal path and a reverse reference path. During training, minimizing its mean reduces a path-space KL divergence and controls an upper bound on endpoint mismatch. During evaluation, the same quantity defines weights for self-normalized importance sampling on paths (path-SNIS), estimates normalizing constants or free-energy differences, and gives the acceptance ratio for path-space independent Metropolis-Hastings (path-IMH). The same forward-reverse laws also define a shared-bridge round-trip Metropolis kernel that acts directly on configurations and preserves the Boltzmann target. On double-well and finite-volume lattice $\phi^4$ targets, the NHMC construction gives corrected estimates when path overlap is sufficient; when overlap is poor, weight degeneracy, low acceptance, and long autocorrelation expose proposal failure. We additionally report a molecular internal-coordinate feasibility study using an MD prior and learned-force path proposal.