arXiv:2109.00586v2 Announce Type: replace
Abstract: The light absorption of [001] grown single-crystalline silicon wafers can be enhanced by chemical etching, e.g., with potassium hydroxide, resulting in a pyramid-like surface texture. Alongside advantageous photon harvesting in solar cells, the surface roughness leads to drawbacks when measuring diffusion behaviour of dopants in the heterogeneous structure. In this paper, we employ experimental and simulated scanning transmission electron beam induced current in combination with simulation of boron diffusion in a self-consistent framework to trace the dopant distribution underneath the pyramid-like surface texture. In order to account for surface recombination, an effective model projecting the system along the electron beam propagation direction is used in the EBIC simulation enabling a comparison to entire two-dimensional experimental maps. We find a good agreement between simulated and experimental data and thoroughly discuss how EBIC can be used in future experiments to quantify weak electric fields.
Science Journals
arXiv:2112.06362v5 Announce Type: replace
Abstract: We consider the problem of scheduling in multi-class, parallel-server queuing systems with uncertain rewards from job-server assignments. In this scenario, jobs incur holding costs while awaiting completion, and job-server assignments yield observable stochastic rewards with unknown mean values. The mean rewards for job-server assignments are assumed to follow a bilinear model with respect to features that characterize jobs and servers. Our objective is to minimize regret by maximizing the cumulative reward of job-server assignments over a time horizon, while keeping the total job holding cost bounded to ensure the stability of the queueing system. This problem is motivated by applications requiring resource allocation in network systems.
A central challenge is to control the tradeoff between reward maximization and fair allocation for the stability of the underlying queuing system (i.e., maximizing network throughput). To address this challenge, we propose a scheduling algorithm based on a weighted proportional fair criteria augmented with marginal costs for reward maximization, incorporating a bandit algorithm tailored for bilinear rewards. Our algorithm admits a regret--queue length tradeoff. For any fixed control parameter $V>0$, it ensures a uniform expected queue length and time-average holding-cost bounds. For a target horizon $T$, choosing $V_T=\Theta(\sqrt{IT})$ at initialization yields $\widetilde O((\sqrt I+d^2)\sqrt T+1/\delta)$ regret. Under this regret-optimized tuning, the corresponding expected queue length and time-average holding-cost bounds remain uniform over the execution time and scales as $O(\sqrt{IT}+1/\delta)$ and $O(\sqrt{IT}/\delta)$, respectively.
arXiv:2204.04887v3 Announce Type: replace
Abstract: Since the era of big data, the Internet has been flooded with all kinds of information. Browsing information through the Internet has become an integral part of people's daily life. Unlike news data and social data on the Internet, cross-media science and technology information data has different characteristics. This data has become an important basis for researchers and scholars to track current hot spots and explore future directions of technology development. As the volume of science and technology information data becomes richer, traditional science and technology information retrieval systems, which support only unimodal data retrieval and use outdated keyword-matching models, can no longer meet the daily retrieval needs of science and technology scholars. Therefore, in view of this research background, it is of profound practical significance to study cross-media science and technology information data retrieval systems based on deep semantic features, in line with domestic and international technology-development trends.
arXiv:2404.03578v3 Announce Type: replace
Abstract: The sim-to-real gap, which represents the disparity between training and testing environments, poses a significant challenge in reinforcement learning (RL). A promising approach to addressing this challenge is distributionally robust RL, often framed as a robust Markov decision process (RMDP). In this framework, the objective is to find a robust policy that achieves good performance under the worst-case scenario among all environments within a pre-specified uncertainty set centered around the training environment. Unlike previous work, which relies on a generative model or a pre-collected offline dataset enjoying good coverage of the deployment environment, we tackle robust RL via interactive data collection, where the learner interacts with the training environment only and refines the policy through trial and error. In this robust RL paradigm, two main challenges emerge: managing distributional robustness while striking a balance between exploration and exploitation during data collection. Initially, we establish that sample-efficient learning without additional assumptions is unattainable owing to the curse of support shift; i.e., the potential disjointedness of the distributional supports between the training and testing environments. To circumvent such a hardness result, we introduce the vanishing minimal value assumption to RMDPs with a total-variation (TV) distance robust set, postulating that the minimal value of the optimal robust value function is zero. We prove that such an assumption effectively eliminates support shift pathologies for RMDPs with a TV distance robust set, and present an algorithm with near-optimal sample complexity. To demonstrate the breadth of our framework, we extend our algorithm and theory to new robust set formulations and robust Markov games. To illustrate the operational relevance, we apply our algorithm to data-driven robust inventory control.
arXiv:2603.22320v3 Announce Type: replace
Abstract: For decades, physics-based climate models have been used to provide insights for climate decision-making. Their application is, however, constrained by significant computational and technical demands. Machine learning (ML) emulators offer a way to reduce these high computational costs; yet, it remains challenging to use ML emulators effectively in climate research. In practice, climate scientists often bypass emulators altogether, and machine learning researchers frequently develop them as methodological showcases without proving their practical utility. The reasons are diverse, ranging from limited accessibility and a lack of specialized knowledge to broader concerns about the physical grounding of ML methods. Here, we discuss limitations and introduce a framework for guiding emulator development, considering both climate science and machine learning perspectives. We argue that designing easy-to-adopt emulators that address clearly defined tasks and demonstrate their reliability is essential. This offers a promising path towards making machine-learning approaches more relevant and usable for applied climate research.
arXiv:2607.11255v1 Announce Type: new
Abstract: Distributed optical fiber sensing (DOFS) combines the advantages of fiber optic sensors, including flexibility, small size, immunity to electromagnetic interference, and high metrological performance, with the capability to transform a single optical fiber into a continuous sensing element for spatially resolved mechanical measurements. Optical frequency domain reflectometry (OFDR), based on Rayleigh backscattering, enables high spatial resolution DOFS measurements, broadening the range of potential sensing applications. However, OFDR based DOFS remains largely unexplored for biomedical applications, despite the need for sensitive, spatially resolved, and conformable sensing interfaces. This study presents a soft DOFS based mat as a large-area interface for physiological monitoring. A single-mode optical fiber was embedded in a flexible silicone matrix and arranged in a serpentine layout to distribute sensing over the mat surface. With a gage pitch of 2.6 mm, the system provided 2250 sensing sites across the active area at a sampling frequency of 50 Hz. The mat was assessed on six healthy volunteers in a seated nearable configuration on the backrest of a standard office chair. The distributed output enabled two dimensional mapping of the mat response, reflecting back mat mechanical coupling and cardiorespiratory induced perturbations. Respiratory rate and heart rate were therefore estimated and compared with a reference wearable system. The maps revealed physiologically coherent spatial and temporal patterns, while the estimated rates showed good agreement with the reference measurements. These results demonstrate the feasibility of combining large area distributed sensing, spatial mapping, and quantitative cardiorespiratory monitoring within a DOFS based soft nearable interface.
arXiv:2607.09771v1 Announce Type: new
Abstract: The bibliography of scientific papers lists items with variable degree of relevance for the contents of the paper itself. If we could identify the sources, i.e., the works that actually inspired the paper, their citations can help us uncover the genesis of scientific projects and would be more representative of the actual importance of papers and authors than the standard citation counts, when all references are considered. Here we present an analysis of the \textit{backbone of science}, i.e., the network of citations between papers and their sources. The latter are extracted from the full body of papers via Large Language Models (LLMs), which are currently very capable of correctly identifying the context in which a paper is cited. Using two different but related prompts, we find that the LLMs select only a small set of references, not taken at random, and that the resulting backbone networks are quite similar to each other with respect to their in-degree distributions, modularity, transitivity, and degree correlations. Backbone networks have higher heterogeneity in their in-degree distributions, compared to the full network, but the most cited papers are usually the same, with some important exceptions. Citation rankings among authors are also remarkably stable. We conclude that the full citation network, despite its redundancy with respect to the backbones, presents a reliable picture of the relative citation impact of papers and authors.
arXiv:2607.09814v1 Announce Type: new
Abstract: This paper is a continuation of our earlier work, in which, we described a Las Vegas algorithm to solve the elliptic curve discrete logarithm problem. The Las Vegas algorithm reduces the elliptic curve discrete logarithm problem to finding a zero minor in a matrix. Using intersection poset of a hyperplane arrangement, we develop an algorithm to find a zero minor in a rectangular matrix. Our methods are elementary. We discuss the complexity of our algorithm, success probability and provide implementation details. We also provide simulation details. Finding a zero minor in a matrix is also of independent interest.
arXiv:2607.11309v1 Announce Type: new
Abstract: Monte Carlo particle transport codes are powerful tools, but their use requires substantial knowledge of input preparation, execution, and result analysis. In this study, we present a code-side strategy for applying existing AI assistants and AI agents to PHITS. Two complementary sets of AI-ready resources were prepared from manuals, lecture materials, sample inputs, utility information, and developer-curated cautions: a bundled knowledge base for retrieval-augmented generation (RAG)-based assistants and a compact agent reference for direct use by AI agents. The knowledge base was loaded into NotebookLM to provide conversational PHITS support, while the agent reference was combined with PHITS-specific policies and execution rules to enable Codex and Claude Code to edit input files, execute calculations, inspect errors, analyze results, and assist with source-code modification and compilation. Five demonstration tasks covered input modification, repeated simulations, parameter optimization, program compilation, post-processing, and result interpretation. The results showed that AI agents could handle complex PHITS workflows when appropriate resources and rules were provided. Practical lessons included precise prompts, human verification, well-documented sample files, explicit execution policies, and command-line-accessible tools. These findings support bundling AI-ready resources with particle transport codes to enable the use of general-purpose AI tools without requiring dedicated code-specific applications.
arXiv:2607.10089v1 Announce Type: new
Abstract: We study the fair allocation of $m$ indivisible items to $n$ agents with additive utilities. In our setting, each indivisible item may be a good, yielding non-negative utility to some agents, or a chore, yielding negative utility to others. Whilst envy-free allocations may not exist in the indivisible-items setting, envy-freeness can be achieved if some amount of divisible good (i.e., \emph{money}) is introduced. When each item's utility or disutility is bounded by one, we show that a subsidy of at most one dollar per agent suffices to guarantee the existence of an envy-free allocation, and that this bound is tight. Moreover, such an allocation can be computed in polynomial time. Since at least one agent need not receive any subsidy, our results imply that a total subsidy of at most $n-1$ dollars suffices to ensure envy-freeness.
arXiv:2607.11233v1 Announce Type: new
Abstract: Virtual try-on (VTON) is a bi-conditional image generation problem that requires not only accurate person preservation but also faithful garment deformation and detail synthesis. Diffusion-based VTON methods can jointly model these factors in a compressed latent space, but suffer from high-frequency detail loss due to inherent latent compression, even with costly multi-step denoising. Recent visual autoregressive (VAR) models offer a promising alternative for high-quality generation with faster inference, yet remain unexplored for VTON due to the lack of effective bi-conditioning mechanisms. To bridge this gap, we first introduce VAR-VTON, a VAR-based VTON model that incorporates garment conditioning and structural guidance for efficient latent-space VTON. Despite its efficacy, latent-space generation still struggles to preserve fine-grained garment details. We argue that different VTON sub-tasks should be addressed in different representation spaces: structural synthesis such as garment warping and person layout is suited to the latent space, whereas fine-grained detail recovery should be tackled in the pixel space. Motivated by this insight, we further propose STAR-VTON, a Two-Stage AutoRegressive framework that builds upon VAR-VTON by decoupling latent-space structural synthesis from pixel-space detail recovery. Our idea is to resort to a matching-informed refiner to establish dense correspondences between the stage-one generation and the source garment to directly map fine-grained pixel-space details. Extensive experiments show that STAR-VTON achieves an impressive efficiency-fidelity trade-off: VAR-VTON runs at least $4\times$ faster than diffusion-based counterparts without degrading quality, and the pixel-space refiner effectively restores fine details and acts as a plug-and-play module that can benefit existing VTON approaches.
arXiv:2607.11240v1 Announce Type: new
Abstract: Ethereum's Glamsterdam upgrade introduces enshrined proposer-builder separation (ePBS), replacing relay-centric PBS with direct builder bids to proposers. We study how this shift changes the block-building microstructure through a general imperfect-information two-stage auction with verifiable messages, where an early bid serves as both a price offer and a signal. PBS and ePBS are modeled as restrictions of the same block-building game: PBS fixes stopping and disclosure exogenously, while ePBS lets the proposer choose stopping and disclosure ex post. Latency heterogeneity is captured by asymmetric information updates: fast builders observe disclosed early information before rebidding, while slow builders do not. We combine exact perfect Bayesian equilibrium characterizations in tractable cases with calibrated no-regret learning in finite games. For PBS, we show that separating equilibria preserve the standard first-price-auction payoff benchmark and provide conditions for their existence. For ePBS, we demonstrate a ratchet effect: because the proposer can defer block proposal and use early bid information in the second stage, builders anticipate ex-post extraction and shade or pool early bids, generating allocation inefficiency and revenue-efficiency valleys.
We interpret this ratchet distortion as a commitment failure. Under full commitment, the optimal policy collapses to the static Myerson auction and removes the ratchet channel. To realize part of this commitment advantage in a feasible mechanism, we propose a Trusted Execution Environment (TEE) sidecar that enforces limited commitment. We formulate the revenue-maximizing TEE mechanism as a bilinear optimization problem. In conservative finite benchmarks, the TEE design increases the proposer revenue relative to the first-price benchmark by approximately \(25\%\).
arXiv:2607.10422v1 Announce Type: new
Abstract: The recent expansion of the FIFA World Cup to 48 teams has prompted discussions regarding a potential further increase to a 64-team format. Scaling the traditional tournament architecture (a round-robin group stage followed by a knockout phase) to 64 teams exacerbates existing structural flaws, notably increasing the frequency of matches lacking competitive relevance and reducing the probability of fixtures between top-ranked contenders. This paper investigates alternative tournament designs by analyzing double-elimination structures for a 64-team mega-event. We evaluate the proposed formats based on competitive fairness, match quality, and scheduling feasibility. Our analysis demonstrates that a double-elimination format eliminates mathematically irrelevant matches and significantly increases the frequency of high-profile games. However, these benefits introduce complex operational constraints, including heightened scheduling complexity and an asymmetric distribution of matches per team, which require specific logistical adjustments. Ultimately, our findings suggest that the continuous scaling of mega-sporting events necessitates a paradigm shift toward non-traditional tournament designs to preserve competitive integrity.
arXiv:2607.10427v1 Announce Type: new
Abstract: Local motion blur detection requires pixel-level localization of blurred regions. Existing benchmarks let models rely on gradient shortcuts that fail to transfer. We introduce BOCCHI (Blurred Objects Captured across Cameras with Human-annotated Imagery), a real-captured benchmark whose sharp regions overlap the blur gradient distribution and defeat these shortcuts, and propose MSDCT-UNet (Multi-Scale Discrete Cosine Transform UNet), a frequency-aware encoder-decoder injecting multi-scale DCT priors through DCT Attention and FiLM. MSDCT-UNet ranks first in in-domain mIoU and boundary localization on BOCCHI, and BOCCHI-trained models outperform every other training source on cross-dataset transfer with only 633 training images.
arXiv:2607.10621v1 Announce Type: new
Abstract: Front-end development accumulates change after change at the repository level, weaving complex cross-file dependencies that current LLM coding agents tuned for single-shot tasks cannot reliably track across multiple iterations, leading to functional regressions and code that resists maintenance. We argue the missing piece is design knowledge: architectural principles, module responsibilities, and structural constraints that developers lean on to keep code readable, maintainable, and evolvable as a system scales. To operationalize this, we propose WebDesignIter, a framework built around a persistent knowledge graph (WebAppArchKG) that fuses repository structure with design knowledge and keeps both in sync across development cycles. WebDesignIter works in two stages: design-informed planning pulls historical context and architectural overviews from WebAppArchKG to produce an implementation plan with corresponding test scripts, and design-aware generation executes that plan through targeted diff-based patches, validated by sandbox execution and automatic syntax repair. On Web-Bench, WebDesignIter delivers an average Pass@2 gain of 9.55 percentage points across nine foundation models over existing baselines. More importantly, WebDesignIter outperforms every general-purpose coding agent Claude Code, OpenHands, SWE-Agent, Codex CLI on every model configuration, posting the highest Pass@1 and Pass@2 while consuming 2530 fewer input tokens. Ablation singles out design knowledge as the most impactful component: stripping it drops Pass@1 by 11.40 percentage points, a degradation far larger than removing code-graph retrieval, patch-based generation, or sandbox verification, confirming that design knowledge provides a fundamentally more efficient and reliable path to repository-level code generation.
arXiv:2607.10810v1 Announce Type: new
Abstract: Deep generative models are increasingly used as simulators for downstream decision-making under data scarcity, but in risk-sensitive applications their usefulness depends on rare adverse scenarios rather than typical samples. Standard generative objectives prioritize bulk distributional fidelity, leaving low-probability tails vulnerable to localized optimization noise and making tail-dependent functionals unstable under finite simulation budgets. We introduce Diachronic Sample Integration (DSI), a test-time inference framework that ensembles generated samples across checkpoints from a stochastic training trajectory. DSI targets a checkpoint-mixture distribution that averages checkpoint-specific tail fluctuations rather than relying on a single brittle endpoint. We formalize this mechanism through a finite-budget bias-variance theory. Empirically, across multivariate synthetic processes and high-frequency trading data, DSI substantially reduces tail-estimation error compared to single-checkpoint baselines under fixed simulation budgets, outperforming standard diffusion and state-of-the-art tail-aware baselines without modifying the generative objective.
arXiv:2607.09834v1 Announce Type: new
Abstract: As human reliance on information technology (IT) increases, having a clear, precise, and comprehensive understanding of the nature of digital objects, digital systems, and digitalized systems becomes more critical. Otherwise, our ability to research, manage, control, and make reliable predictions about them will be limited. Accordingly, in this paper, we propose a new theory of digital objects, digital systems, and digitalized systems based on the adoption, adaptation, and extension of existing theories of ontology, semantics, and semiotics. The theory provides precise explanations of the nature of digital objects, digital systems, and digitalized systems. Ours is a realist theory that does not countenance the independent existence of nonmaterial or hybrid objects in the world. Accordingly, we are at odds with much of the prevailing discourse about digital phenomena. We show how our theory generates different insights and predictions from this the dominant discourse on the nature of the digital. Our predictions lay the groundwork for further empirical studies on the design, use, and impact of IT on individuals, organizations, and society.
arXiv:2607.11417v1 Announce Type: new
Abstract: Advanced physics laboratories require students to integrate disciplinary knowledge, experimental practice and scientific argumentation across complex observational and analytical tasks. The increasing availability of generative artificial intelligence (GenAI) adds complexity to this coordination, since AI systems may function as conceptual explainers, operational assistants, artefact reviewers or apparently authoritative evaluators. This exploratory qualitative case study examines AstroTutor, a constrained GenAI tutor introduced as an optional support resource in a Master's-level advanced astrophysics laboratory. The study investigates how students framed the tutor within a broader GenAI-mediated learning ecology that included the instructor, peers, course materials, observations, measurements, data analysis and final assessed reports. Seven students attended the course, five used the tutor, and three groups produced a final report. The analysis combined content analysis, thematic analysis and frame analysis. Drawing on chat logs, final reports and limited post-use reflective responses, the results identify five principal GenAI functions: interface interpreter, warrant organiser, report scaffold, unstable authority and resource whose traces may appear in downstream reports. These findings extend previous research on GenAI in education to the context of advanced physics laboratories, showing that its use requires explicit design boundaries, guidance on legitimate and prohibited practices, verification routines, and assessment requirements that preserve students' epistemic responsibility. The educational implications of a GenAI-mediated learning ecology in advanced physics laboratories are also discussed.
arXiv:2607.11563v1 Announce Type: new
Abstract: Opal2 self-encrypting drives provide hardware-based disk encryption serving as an additional layer of protection, or a replacement, for software-based solutions. This paper presents a case study of real-world Linux integration of Opal2 drives and the security of Opal2 firmware. The study was conducted on a testbed of 38 commercial off-the-shelf Opal2 drives from various vendors using a black-box approach. We identified several firmware security issues and incompatibilities, which we responsibly disclosed to respective vendors. Our findings led to improvements in Linux disk encryption tools used across all major Linux distributions. To enable independent evaluation for the public, we release our test scenarios for Opal2 drives as an open-source toolset.
arXiv:2607.11229v1 Announce Type: new
Abstract: The local failover routing is a mechanism that routes a packet from a source to a destination only using pre-calculated routing tables, even when several edges fail. In this paper, we study local failover schemes that minimize the number of rewritable bits in the packet header on directed graphs with $k$-arc failures. There are many studies of failover routing on undirected graphs, and it has been investigated whether routing is possible depending on the number of bits in the packet header, the type of failure, the graph properties, etc. In contrast, there is not much research on directed graphs. Van et al.~first showed the upper and lower bounds of rewritable bits in the packet header on directed graphs. However, their results showed a large gap between the upper and lower bounds. The main contribution of this paper is to close the gap between the upper and lower bounds. Specifically, we show that our scheme can route packets with $k$ faulty arcs if the packet header has $\min(k \log ( \frac{e(2n+k-3)}{k}, 2n \log ( \frac{e(2n+k-3)}{2n})))$ rewritable bits, where $n$ is the number of nodes. Moreover, any local failover routing scheme needs $\Omega(k\lceil\log\frac{n}{k}\rceil)$ rewritable bits when the number of faulty arcs is equal to or less than $\frac{3(n-1)}{8}$ and $\frac{n-1}{4}$ rewritable bits when the number of faulty arc is more than $\frac{3(n-1)}{8}$. This result means our scheme is nearly optimal when the number of faulty arcs is approximately less than the number of nodes.
arXiv:2607.10156v1 Announce Type: new
Abstract: Microservice traces can be structurally anomalous even when every span returns normally -- a payment flow that silently skips a risk check looks fine to any per-span monitor. Sequence models like DeepLog address this by predicting the next event, but they treat each API endpoint as a context-free token: the same endpoint reached through different invocation chains is mapped to the same vocabulary entry, even when its normal behavior differs across contexts. We propose encoding each event as an (endpoint, root-to-span invocation chain) pair instead. This simple change has two consequences: unseen chains are flagged without model inference, and next-event predictions become context-conditional, turning subtle path anomalies into clear outliers. We instantiate this idea in CHAINLSTM, a lightweight dual-task LSTM supporting per-event online detection. On the TrainTicket benchmark, CHAINLSTM achieves 94.3% F1 (+5.3 pp over DeepLog) with comparable latency recall and 99.1\% path recall. Case analysis shows that chain-aware encoding shifts median prediction probability on path anomalies from 0.91 to 0.002, suggesting a wider separation margin for threshold-based detection.
arXiv:2607.11458v1 Announce Type: new
Abstract: Multi-aperture coherent reception can improve freespace optical (FSO) links by converting spatial diversity into coherent combining gain. In turbulent links, the aperture branches are simultaneously affected by relative phase errors, polarization mismatch, and unequal signal-to-noise ratios (SNRs). Existing methods treat phase/polarization alignment and branch-weight allocation as separate operations, or absorb all impairments into a high-dimensional MIMO equalizer that obscures the physical meaning of each aperture's contribution. This paper proposes a structured blind combining method based on heterogeneous gradient sources: phase and per-aperture polarization parameters are updated by closed-form analytical gradients that maximize the combined output power, while aperture weights and an optional global polarization angle are updated by gradients derived from the constellation-radius error. An exponential parameterization pn = eqn/N ensures positivity without clipping. The internal variable qn is adapted by radius-error gradients, thereby allocating maximal-ratio-combining-like weights according to the quality of the already aligned branches.
arXiv:2607.11573v1 Announce Type: new
Abstract: Large language models (LLMs) have opened new opportunities for unit test generation, but executable tests do not necessarily reveal real defects. This paper studies how historical real-bug mechanisms can be transformed into executable feedback targets for LLM-based unit test generation. The proposed framework constructs structural and semantic representations of real-bug records, retrieves mechanisms applicable to a focal method, and instantiates them as synthetic bugs that guide iterative test enhancement. We evaluate the approach on method-level real-bug detection tasks from Defects4J and show that mechanism-guided synthetic-bug feedback improves real-bug detection over execution-, coverage-, mutation-, knowledge-, and search-based baselines. The results suggest that organizing real-bug mechanisms as retrievable and executable feedback targets is an effective way to guide generated tests toward bug-triggering inputs and behavioral oracles.
arXiv:2607.10354v1 Announce Type: new
Abstract: We develop a sketch-and-restart framework for computing the action of a matrix function on a vector, $f(A) b$, where $A$ is large, sparse, and non-Hermitian. The framework combines quadrature-based restarting with Arnoldi-like decompositions generated by sketched or truncated Arnoldi processes. Within this framework, we develop two classes of restarted algorithms. The first uses a fixed Krylov subspace dimension and is based either on the sketched Arnoldi process or on a new sketched harmonic Arnoldi process proposed in this work. The second class chooses the Krylov subspace dimension adaptively by running the truncated Arnoldi process until the condition number of the generated basis, estimated from its sketch, exceeds a prescribed threshold. We also establish the convergence of the restarted sketched harmonic Arnoldi method for Stieltjes functions under the assumption that $A$ is positive real. Numerical experiments demonstrate the effectiveness of the proposed framework, including the computational savings achieved through sketching, the storage reduction enabled by adaptive truncation, and the acceleration obtained from thick restarting.
arXiv:2607.11586v1 Announce Type: new
Abstract: Mixture-of-Experts (MoE) large language models (LLM) activate only a small number of experts during inference, but token routing introduces persistent expert hotness skew: a small set of hot experts continuously receives most tokens, while the remaining experts are lightly loaded. On 3.5D multi-chiplet systems, this skew not only causes compute imbalance but also amplifies pressure on communication, memory bandwidth, I/O, and execution queues. Therefore, the core problem is not simply to reduce token movement, but to dynamically place and reuse hot expert replicas across different memory tiers.
This paper proposes HCRMap, a hot expert residency mapping framework for pressure-aware expert replica management in 3.5D MoE inference. Based on expert hotness, weight loading cost, migration overhead, and runtime resource pressure, HCRMap dynamically determines which experts should be promoted, retained, demoted, or evicted. It then maps routed token groups to suitable resident replicas, thereby jointly mitigating communication, memory, and queue bottlenecks.
Experimental results show that HCRMap reduces end-to-end latency by 43.6% and 43.0% over Hydra in the prefill and decode stages, respectively; by 34.5% and 33.1% over MoEntwine; and by 46.7% and 46.0% over PIMoE.