Papers

Sampling-based test-time scaling enhances LLM reasoning by aggregating multiple reasoning paths for the same query. Existing methods use answer frequencies or path confidence to compute candidate scores, which do not explicitly characterize comparisons between competing answers. Inspired by the Condorcet principle in social choice theory, we introduce pairwise confidence comparisons between paths supporting different answers to capture answer-level comparative advantage. We combine this comparative evidence with voting support through a KL-regularized objective. Theoretically, we show that this comparative-advantage is equivalent to the adjustment to the traditional voting logits. In this way, our proposed Pairwise-Guided Logit Adjustment (PGLA) framework allows comparative evidence to compensate for weaker voting support and correct voting-based ranking errors. Experiments on 10 challenging benchmarks including mathematical reasoning, knowledge reasoning, and code generation with 7 models ranging from 1.7B to 32B parameters speak to the effectiveness of this plug-and-play term. Additional evaluations with external reward scores and online sampling further demonstrate its applicability.

We investigate how unresolved quasistatic disorder in system bath coupling affects non-Markovian dynamics in spin environments. The coupling is fixed within each experimental shot but varies between shots, so repeated shot tomography reconstructs a disorder averaged dynamical map. For a single spin bath with pure dephasing Ising interaction, disorder can drive a transition to CP-divisible dynamics for fixed bath preparations, with a critical strength set by bath polarization along the interaction axis. However, when the bath state fluctuates physically between shots, the averaged dynamics remain CP indivisible at any finite disorder strength. We extend these results analytically to inhomogeneously coupled pure dephasing baths of any number \(N\) of non-interacting spins. At any finite disorder strength, one unpolarized bath spin suffices to retain CP-indivisible dynamics, whereas a fully polarized bath yields Markovian (CP-divisible) dynamics. Dynamics with physically fluctuating bath states likewise remain non-Markovian (CP indivisible) at finite disorder for arbitrary \(N\) in the pure dephasing case. For anisotropic Heisenberg interactions, any nonzero transverse exchange eliminates the disorder driven transition to CP divisibility in a single spin bath. Exact diagonalization of anisotropic models with inhomogeneous couplings up to \(N=12\) reveals persistent trace distance revivals for fixed bath preparations, sampled Haar random preparations, and physically fluctuating bath states, showing that this effect of transverse exchange extends beyond a single bath spin. Increasing disorder still suppresses experimentally accessible revivals and can make memory effects unresolvable at finite measurement resolution even as the dynamics remain CP indivisible. Thus, transverse exchange and physical bath state fluctuations preserve dynamical indivisibility against unresolved coupling disorder.

A seller without commitment to future offers interacts with a buyer over time. There are many goods for sale, with the buyer's value for a bundle the sum of each bundled good's value. Each date, the seller jointly announces a menu of subsets of the remaining inventory and current prices, and the buyer may purchase one of the offered subsets or wait. In the next period, the seller may again offer any menu of the remaining goods. Neither buyer nor seller can commit to future behavior. I show that as players become maximally patient, if scope increases sufficiently fast compared to patience, revenue per good converges to expected value per good and buyer utility per good converges to zero, in any equilibrium. But if patience instead grows sufficiently fast compared to scope, seller revenue per good converges to the lowest possible value and the buyer captures all remaining surplus, in any equilibrium.

We introduce and analyze a two-level overlapping additive Schwarz preconditioner for discontinuous Galerkin discretizations of virtual element spaces on general polygonal meshes in two dimensions, for a second-order elliptic problem whose coefficient is constant on each subdomain and may have large jumps across the interface. The discretization is a weighted symmetric interior penalty formulation: the averages carry the coefficient and the penalty is scaled by its harmonic mean across each edge, which keeps the method consistent and makes the constants of the analysis independent of the coefficient. The coarse space is obtained from a virtual, continuous partition of unity with a single degree of freedom per subdomain vertex, and is therefore applicable to the irregular subdomains produced by a graph partitioner; the discrete harmonic extension used to build it may be replaced by a cheaper projection onto polynomials of low degree. We prove that the condition number of the preconditioned system is bounded by $C(1+\log(H/h))(1+H/\delta)$, where $H$, $h$ and $\delta$ denote the subdomain diameter, the mesh size and the overlap, and where $C$ is independent of the number of subdomains and of the contrast of the coefficient, provided the latter is quasi-monotone. This is the same bound known for the conforming case and the analysis allows for hanging nodes. We also show, by an explicit example, that the quasi-monotonicity assumption cannot be dispensed with. Numerical experiments with structured triangular, hexagonal and Voronoi meshes, and with subdomains obtained both by uniform partitioning and by METIS, confirm the theory.

System One models choose an agent's next action by scoring a supplied set of options in a single pass. This makes them fast and cheap, but when they choose wrongly there is nothing to inspect: did the model misread what the agent has seen, or read it correctly and still act wrongly? We answer behaviorally. On a frozen benchmark of agent decision states from simulated environments, customer-service policies, tool use, web navigation and coding, we pair every next-action question with factual probes about the same state, each answered independently. Across seventeen fast systems and a reasoning model, the answer is consistent: most wrong actions follow correct reading, and the more accurate a system, the more of its errors take this form. These errors are not noise; they gather on the same states across systems of every kind. Where the probed facts provably determine the action, models that read every fact still fail to combine them, and their own answers, composed by code, repair the failure. On real states, what is missing is often not a fact but the next sub-goal.

Rate splitting multiple access (RSMA) has emerged as a promising non-orthogonal transmission mechanism to improve spectral- and energy efficiency, and user fairness in multi-antenna broadband channels. To further enhance its performance, in this paper we investigate a novel cooperative RSMA (C-RSMA) transmission framework assisted by reconfigurable intelligent surface (RIS) and user relaying. More specifically, we first formulate the common- and private-rate maximization problem satisfying the transmit power constraint, by jointly optimizing the transmit beamformers at the BS and the reflective beamformer at the RIS. The above formulated problem is an intractable and NP-hard problem as the optimization variables are coupled and the strong residual self-interference (SI) is involved. To conquer the challenge mentioned above, we propose a low-complexity alternative optimization approach to jointly optimize the transmit and reflective beamforming. Simulation results show that our approach is capable of enabling superior sum-rate performance than the RIS-assisted space division multiple access (SDMA), half-duplex RSMA, and SDMA without using RIS schemes.

Best-arm identification (BAI) aims to identify the optimal arm using as few samples as possible. While most existing work focuses on a single objective, many real-world applications involve multiple criteria, which motivates the study of BAI in multi-objective bandits. Intuitively, multi-objective identification seems harder, since the learner must identify optimal arms in a higher-dimensional reward space; however, the additional objectives also provide richer feedback that can reduce the required exploration. In this paper, we characterize when and why multiple objectives can make BAI easier. For Gaussian bandits, we establish a characteristic-time lower bound for fixed-confidence identification that holds uniformly across preference orders. Through explicit comparisons under Pareto, linear, and lexicographic preferences, we reveal three phenomena: (i) identifying a larger optimal set can require fewer samples than identifying a unique optimum; (ii) the advantage of vector feedback over coordinate feedback depends fundamentally on the preference order; and (iii) multi-objective identification can require fewer samples than even the easiest corresponding single-objective problem. Finally, we propose Order-Aware Track-and-Stop (OATS), an algorithm that is correct at any prescribed confidence level and asymptotically matches the lower bound.

The massive disk galaxies (stellar mass >= 10^11 M_solar ) in the local universe (z < 0.1) encode challenging aspects of galaxy formation physics. We have studied one class of such galaxies - the giant low surface brightness galaxies (gLSBs) that show a hybrid morphology - a central Sersic, an inner high surface brightness, exponential stellar disk surrounded by an outer, extended low surface brightness disk, using large volume IllustrisTNG-100 simulation. This is motivated by the recent observation of several such systems and an earlier work by Sarkar & Saha (2025) from IllustrisTNG-50. We find that despite using a ~ 10 times larger volume simulation, only ~11% of massive disk galaxies show such morphology, consistent with findings from TNG50. This shows robustly that such galaxies are not common in the local universe. We compared them with the typical disk galaxies from observations, and single-exponential gLSBs from TNG100. We show that while the inner disk components lie well within the central surface brightness vs. radial scale length distribution of the typical disks, the outer LSB disks lie significantly at the outskirt, occupying a unique space. Their median bulge-to-total ratio is ~ 0.2 making them disk dominated, similar to the typical disks and the single-exponential gLSBs. We find their median dark matter mass fraction to be high (~ 0.9), typical of gLSBs. The outer LSB disks reach fainter central surface brightness and larger scale length values than the single-exponential gLSBs. Our exploration will be useful in understanding their formation mechanism across cosmic time.

Post-training INT8 quantization is typically validated by checking accuracy retention, which requires ground-truth labels often unavailable in low-resource clinical settings. We show a label-free alternative: comparing a model's own output before and after quantization. We compute cosine similarity between the FP32 and INT8 models' output probability vectors for each prediction, which requires only a single forward pass. This output-agreement similarity separates lower-accuracy predictions across four evaluated architectures (ResNet-50, VGG-16-BN, Wide ResNet-50, and ViT-B/16) under fixed-percentile splits, outperforming a confidence-only baseline for all four. These results show that quantized models can be compared to their baselines to assess their drift and determine whether they are well suited for clinical deployment.

Let $\tau(G)$ denote the maximum number of pairwise edge-disjoint spanning trees contained in a graph $G$. Zhang and Fan [J. Graph Theory, 2026] posed the problem of determining the corresponding tight spectral-radius condition for every $s$-edge-connected graph with \[ k+1\leq s\leq2k-1, \] together with a characterization of the associated extremal graphs. In this paper, we completely resolve this problem.

The stories told about artificial intelligence (AI) play a central role in shaping public understanding, policy debates, and expectations about technological change. News media are especially influential in this process. Using a cross-regional corpus of 12,924 English-language news articles about AI published since 2022, we examine how storytelling about AI varies across media systems in four regions: North America, China, India, and the Middle East. Leveraging large language models, we extract the implicit narrative lessons conveyed by each article—commonly known as story morals—capturing value-laden interpretations that extend beyond thematic content or sentiment. We find that North American AI coverage is uniquely characterized by narratives of caution, while under-emphasizing narratives of innovation and collaboration that are more prevalent elsewhere. However, this pattern is largely explained by a broader regional negativity bias in North American news sentiment, rather than by heightened concern about AI itself. Our findings bring into view the culturally conditioned nature of AI news narratives.

Automated post-training transforms pre-trained large language models (LLMs) into models capable of handling diverse domains by automatically constructing and optimizing a recipe of coupled decisions over data sources, data processing, cross-domain composition, and training configuration under a limited experimental budget. Existing automation either searches predefined recipe spaces that enable systematic optimization but remain fixed rather than adapting their available choices to the target tasks, or uses open-ended agents that discover task-specific choices but leave them implicit in trial-level edits, limiting their systematic reuse and recombination across trials. To fill this gap, we introduce optimization over a task-adapted recipe space, a novel paradigm that can adapt the recipe space itself to the target tasks while representing its choices explicitly for systematic reuse, recombination, and joint optimization. We instantiate this paradigm with Structured Recipe Optimization (SRO), comprising an LLM-based Recipe-Space Designer that constructs the task-adapted recipe space and a Multi-Domain Recipe Optimizer that uses structured trial evidence to guide LLM candidate construction and multi-objective Bayesian trial selection. Extensive experiments show that SRO achieves the highest average score in all evaluated target-model and agent-LLM combinations. In joint post-training across three capability domains and six benchmarks, SRO reaches 41.1 versus 15.4 for the strongest baseline on Qwen3-1.7B with GPT-5.4, and 53.8 versus 40.8 on Qwen3-4B with DeepSeek-V4-Flash. Our code is available at https://github.com/RuohLiuq/SRO.

Different MRI modalities reveal different parts of a brain tumor. Segmentation models combine information from these modalities to outline tumor regions. Many models use a separate encoder for each modality. However, exchanging features among these encoders at several stages can increase model complexity. We introduce the Block-Wise Shared Adapter (BWSA), which combines compressed features from the modality encoders in a shared path and returns adapted features to each stream. In its fixed-blending form, BWSA adds the adapted features to the original features. Its trainable-gated variant (TG-BWSA) instead learns separate weights for the original and adapted features at each modality and encoder stage. We evaluate both variants on BraTS 2020 and BraTS 2023 with four segmentation backbones. On the official BraTS 2023 online validation set, BWSA increases average region-wise Dice by 1.3–7.6 percentage points relative to the corresponding unmodified backbones. TG-BWSA increases it by 2.8–7.7 points. Both variants add approximately 0.85–3.73% parameters and reduce average HD95 relative to the unmodified backbones.

Latent planners often rank candidate action sequences by terminal Euclidean distance to a goal latent. Exact agreement with transition distance would make this score ideal, but is stronger than planning requires: along a transition-shortest rollout, the essential requirement is that latent distance does not reverse true transition progress. We show that this pathwise no-reversal requirement is precisely the classical self-approaching condition. To the best of our knowledge, we are the first to identify self-approachingness as a planner-facing compatibility condition for terminal Euclidean latent planning and to connect its violation to an exact finite-horizon candidate-order reversal. Because visual world models predict discrete latent sequences, we analyze their polygonal interpolation. We prove that an acute bend creates a nonempty region in which terminal Euclidean distance ranks a transition-worse candidate ahead of one that makes genuine progress, even under exact rollouts. We further interpret latent dimension as angular capacity through a classical spherical-code bound. A controlled six-branch experiment, including a zero-padded three-dimensional baseline, validates the predicted ordering failure and separates angular arrangement from ambient dimensionality.

Vision language models (VLMs) are used in creative tasks such as visual storytelling, design ideation, art interpretation, and advertising. Yet we lack principled ways to measure the lateral reasoning of these models compared to human potential. We introduce CREATE, a semi-automatically constructed benchmark containing 155 visual Remote Associates Test problems. These problems require a solver to find an underlying connection among three seemingly dissimilar images. We conduct a human study (N=25) to validate and ground the quality of our benchmark, and evaluate 11 VLMs. Our findings show that most VLMs underperform humans on this task. Our experiments indicate that the associative ability of VLMs depends on the cue concepts provided in the text and does not universally extend to visual puzzles.

Vision-language-action (VLA) policies excel at generalist robotic manipulation but generate actions without explicitly modeling their environmental consequences. World action models (WAMs) incorporate future prediction, yet most rely on costly video generation, while latent alternatives either use predicted futures only as auxiliary targets or infer latent actions from suboptimal visual transitions through costly large-scale video pretraining. We introduce WAVE-0 (World-model-Augmented VLA Execution), a framework that bridges VLAs and WAMs through a lightweight latent world model. By connecting the semantic understanding of pretrained VLAs with explicit future prediction, WAVE-0 allows anticipated action consequences to directly guide action generation. Built on a pretrained VLA, its 257M world model translates VLM-generated latent action plans into predicted future representations that condition the action expert, unifying latent planning, future prediction, and action generation. The world model is trained directly on ground-truth robot actions with a counterfactual ranking objective, requiring neither pixel reconstruction nor large-scale video pretraining. WAVE-0 achieves success rates of 91.47% on RoboTwin 2.0 and 98.7% on LIBERO, along with 88.0%, 61.4%, and 75.0% on LIBERO-Plus, LIBERO-Pro, and LIBERO-Para, respectively. On real-world tasks, it improves the success rate of $\pi_{0.5}$ from 38.3% to 73.3%. These results demonstrate strong manipulation performance and generalization across diverse distribution shifts.

Retrieval-Augmented Generation (RAG) has emerged as a dominant paradigm for improving Large Language Model (LLM) by intertwining it with other components, such as a vector/graph database. Therefore, the resulting RAG systems are often highly configurable, under which a non-trivial amount of failures are related to the configurations. However, there remains a lack of systematic understanding regarding the configuration issues for RAG systems. To bridge this gap, in this paper, we conduct the first large-scale empirical study of 654 real-world configuration issues collected from 10 diverse RAG systems. Drawing on this, we provide a complete categorization of the symptoms, root causes, relevant stages, and fix patterns of configuration issues for RAG systems, as well as several new observations. Our findings not only reveal the characteristics of configuration issues, but also strongly link them to the unique pipeline stages of RAG, e.g., we found that most symptoms can be observed at the early or late stage, while the stage of the root cause might not be at the same stage as where the symptom is identified. We also provide nine actionable insights to the practitioners on testing, localizing, and fixing the configuration issues for RAG systems.

Large language model (LLM) routers reduce serving cost by choosing which model to call and how many output tokens to grant it for each query. Existing routers rank model-budget actions using a single quality metric. This does not work for mixed workloads. For ordinary queries, a good answer is correct and useful (utility); for security-sensitive queries, a good answer refuses harmful requests or resists malicious instructions (security). These two objectives prefer different models and budgets. We propose SCRouter, a router that learns when to prioritize security and how to allocate the token budget. For each model-budget pair, it estimates the probability that the response finishes within budget. It uses this probability to predict utility and security separately. A small probe determines how to weight these two predictions for each query. SCRouter then selects the model and budget by balancing the weighted utility and security against cost. We also build SCRouterBench, a benchmark of 38,869 utility and security queries spanning nine security dimensions, each answered under all 144 combinations of 12 models and 12 budgets, plus 2,753 out-of-domain queries. SCRouter improves the baselines by 13-29 points in-domain and 19-23 points out-of-domain.

Simulation of HPC applications is an essential tool for analyzing system design trade-offs and enabling effective hardware–software co-design. As modern supercomputers increasingly rely on GPU acceleration, MPI implementations have evolved to support GPU-aware communication, enabling direct GPU data transfers. Modeling these mechanisms is therefore critical for simulating HPC applications on current systems. In this paper, we introduce Orrery, an interposition layer built on SimGrid's SMPI to model GPU-aware MPI communication using existing collective implementations and flow-based simulation. Orrery intercepts GPU memory allocations, kernel launches, data transfers, and MPI operations, and models their execution in SimGrid according to target system specifications, enabling simulation without requiring physical GPUs. We validate Orrery using CUDA-aware MPI microbenchmarks and a Jacobi solver on up to 32 nodes of the MUSICA supercomputer and demonstrate that it reproduces key GPU-aware MPI mechanisms.

Aerial map matching provides position observations for navigation, but a position error alone does not determine how an observation should be weighted. We define five benchmark tasks: position availability and agreement, grouped error structure, predictive uncertainty, controlled updates, and temporal accumulation. A case study in digital-surface-model-assisted localization evaluates base-anchored uncertainty calibration within this framework. The method combines propagated image and map sensitivities with the second moment of a compact configuration subset relative to the reported position. The reported position remains unchanged, and scale is fitted using separate position references without local surface-height ground truth. Thirteen covariance models are evaluated on two held-out query sets from locations represented in the calibration data. Relative to a constant full-matrix baseline, the method reduces equal-location normalized position-update risk by 7.70% on 728 conditions and 5.46% on a separate 142 conditions under an independent prior, with all localization failures included. On the primary set, equal-location negative log likelihood improves by 2.571 nat. These results show improved reference-based weighting at seen locations. Exploratory base-quality and conditional-addition controls leave the incremental update benefit of configuration disagreement unresolved. Navigation benefit and real temporal accumulation remain untested.

Transient stability constraints alter the economic dispatch of power systems, while the resulting price signals and their implications for market settlement remain insufficiently characterized. Existing approaches identify stability-related components in locational marginal prices, without deriving shadow prices for individual assets, such as generators and loads, or analyzing the resulting settlement. To address this gap, this paper derives shadow prices from a transient stability-constrained optimal power flow in which the differential-algebraic equations governing the system dynamics are embedded through direct discretization. Each shadow price decomposes into a nodal energy component and a stability component, the latter formed by the dual variables of the dynamic equations in which the power of the asset appears. Shadow prices are therefore asset-specific: co-located generators and loads are assigned the same nodal energy component, but can be subject to different stability components when their transient behaviors differ. Case studies on the WECC 9-bus system and a modified IEEE 39-bus system show that settlement at the proposed shadow prices allocates the cost of the stability constraints among the assets without any side payment. An asset that reduces the stability margin is charged its stability component, whereas an asset that improves it is paid, and no generator is dispatched at a loss.

Let $X$ be a smooth algebraic variety over $\mathbb{R}$. We study the separation of the connected components of $X(\mathbb{R})$ by equivariant cohomology, unramified cohomology, and signatures of unramified quadratic forms. Our results highlight how the geometry of $X$ influences these questions. We apply them to cubic hypersurfaces, complete intersections of two quadrics, abelian varieties, and Enriques surfaces. We also obtain new bounds on the exponent of the cokernel of higher signature morphisms in $\mathbf{I}$-cohomology and investigate the optimality of these bounds for smooth projective varieties.

Lattice vibrations govern the propagation of heat in materials, and investigating their dynamics through time-resolved phonon dispersion measurements would provide insight into these excitations and their interactions. However, time-resolved measurements that directly resolve the phonon dispersion at the nanoscale have not been previously demonstrated. Momentum-resolved vibrational electron energy loss spectroscopy (EELS) provides access to phonon energies and momenta with nanoscale spatial resolution. Here, we combine momentum-resolved vibrational EELS with pulsed laser heating and synchronised gated electron detection to observe the phonon dispersion of graphite during controlled heating and cooling on the microsecond timescale. Recording both the gain and loss spectra at defined times relative to laser excitation, we determine the temperature at each time delay and follow changes to the phonon branches as a function of temperature. The momentum-resolved spectra reveal a decrease in energy of the optical phonon branch with increasing temperature accompanied by an increase in peak width. These changes are consistent with temperature-dependent phonon renormalisation and scattering associated with anharmonic interactions. The approach is compatible with detector gating on timescales down to 50 ns, providing a route towards extending these measurements to faster dynamics. This work demonstrates time-resolved phonon dispersion spectroscopy using vibrational EELS and provides a route towards probing temperature-dependent phonon dynamics at the nanoscale with simultaneous energy and momentum resolution.

We prove that the last iterate of Optimistic Follow-The-Regularized-Leader (OFTRL) converges at a linear rate to a Nash equilibrium in multiplayer harmonic games. Our result holds under a mild game-regularity condition and for a large family of regularizers, including entropy, Tsallis, and log-barrier. Our analysis is based on a dual perspective of OFTRL: we prove that the interaction between optimism and the global structure of harmonic games leads to the dissipation of an energy function along the dual iterates, for which we establish tight quantitative bounds. Our result also holds for the Extragradient FTRL variant.

Mode locking is the mechanism by which lasers organize many optical modes into stable trains of ultrashort pulses. It underpins optical frequency combs, precision timing and spectroscopy, and provides the phase-coherent waveforms required for applications ranging from strong-field physics to optical frequency metrology. Conventional pulse-forming mechanisms select pulse amplitude without fixing its phase relative to an external optical reference, leaving pulse formation and optical-phase stabilization as separate functions. Phase-sensitive amplification offers a different possibility: the optical reference can participate directly in pulse formation. Here we demonstrate this principle in a continuously pumped oscillator. Kerr-induced phase evolution and cavity detuning turn pump-referenced parametric gain into a power-dependent discriminator that supports stable single-soliton operation. Roundtrip-resolved interferometry reveals that the soliton optical phase remains referenced to the master laser modulo $\pi$. These results establish phase-sensitive mode locking, in which pulse-amplitude selection and optical-phase referencing arise from the same coherent nonlinear feedback.

Region-of-interest (ROI) selective encryption in H.265/HEVC aims to protect sensitive content while preserving the utility of non-ROI regions. Existing coding-unit (CU)-level approaches still suffer from substantial encryption redundancy for two reasons. First, existing approaches map object-detection bounding boxes to CUs, producing regular rectangular regions, which poorly match the shape of irregular ROI boundaries, such as facial contours. Second, their coding-tree-unit (CTU)-based diffusion isolation expands the protected footprint in fixed $64\times64$ blocks, causing encryption-induced distortion to spread over a large number of non-ROI pixels. To address the aforementioned shortcomings, we propose a prompt segmentation-guided CU-level ROI encryption algorithm, which mainly consists of two core components. The first component is a prompt segmentation-guided fine-grained ROI encryption pipeline. It maps pixel-level ROI masks to CUs according to an ROI-overlap threshold, refines boundary CUs to $8\times8$ granularity, and applies format-compliant encryption to multiple syntax elements in the mapped ROI CUs. The second is a dual diffusion isolation strategy to address the encryption diffusion problem. For spatial isolation, the encoder refines non-ROI boundary regions into $8\times8$ CUs and uses PCM mode to bypass intra-frame prediction. For temporal isolation, it removes motion-vector candidates whose reference blocks overlap ROI CUs. Experimental results demonstrate that the proposed method generates encryption regions that more closely match irregular ROI boundaries, effectively confines distortion to the ROI.

Large Language Models (LLMs) execute billions of floating-point operations per output token across many hardware accelerators. At this scale, hardware-induced faults can bypass system-level detection mechanisms and cause Silent Data Corruption (SDC). Algorithm-Based Fault Tolerance (ABFT) is the state-of-the-art technique for protecting matrix multiplication from such faults, but it incurs non-negligible computational overhead and requires thresholds to distinguish hardware-induced errors from floating-point roundoff. In this paper, we revisit Cauchy–Schwarz Norm Bounds (CSNB) as a low-overhead alternative for detecting SDC during LLM inference. We compare CSNB against ABFT through controlled fault injection and evaluate the runtime overhead of both verification and threshold determination across multiple model sizes, batch sizes, and generation lengths. Our results show that, on average, CSNB retains 99% of the fault detection capability of ABFT for the evaluated linear layers in dense LLMs, while reducing verification overhead by 87% and threshold determination overhead by 67%. For Mixture-of-Experts (MoE) router gate computations, however, CSNB retains only 55% of the fault detection capability of ABFT. These findings suggest that CSNB is a promising low-overhead alternative for protecting linear layers in dense LLMs, while checksum-based verification remains preferable for the evaluated MoE router gate.

Deep learning for medical image restoration is often constrained by the need for paired training data. Acquiring perfectly registered clean-noisy radiographs in vivo is impractical due to patient motion and radiation-dose limits. To address this, we employ a self-supervised Noise2Noise paradigm that trains on independently corrupted observations without paired clean images. We propose Denoise X, a lightweight self-supervised U-Net model (7.79M parameters) for suppressing quantum mottle and electronic readout noise in low-dose chest radiographs. It is trained on the NIH ChestX-ray14 dataset (112,120 images) using a stochastic Poisson-Gaussian noise simulator. Zero-shot evaluation on the unseen Montgomery County dataset achieved a PSNR of 36.09 dB and SSIM of 0.8806, outperforming the Bilateral filter while preserving high-frequency anatomical details. Using patch-wise inference, Denoise X processes native-resolution chest radiographs in ~2.2 s per image on a Tesla P100 GPU. Cross-domain evaluation on the Shenzhen dataset achieved 36.44 dB PSNR and 0.9099 SSIM, showing strong generalization. For resource-constrained deployment, the model was converted into a 31.16 MB ONNX model for efficient CPU inference on non-DICOM images.

In this paper, we study the last-iterate convergence of clipped stochastic gradient descent under heavy-tailed noise. We assume that the stochastic gradients are unbiased and that the noise has a bounded central $p$-th moment, where $1<p\le2$. For smooth convex objectives, a linearly decreasing learning rate gives a high-probability error bound of order $O(LR^2\log(1/\delta)/T+\sigma R(\log(1/\delta)/T)^{(p-1)/p})$. We further establish last-iterate guarantees for nonsmooth composite objectives and smooth composite objectives with proximal updates. In addition, when a causal adversary replaces at most $q$ gradients, the same clipped update achieves a bound with $\log(1/\delta)$ replaced by $\log(1/\delta)+q$. The corresponding lower bounds for the clean and corrupted oracle models are also provided.

In this work we incorporate thermal fluctuations into fluid-dynamic simulations with the vHLLE code, in linear approximation. The fluctuations are built in following the fluctuation-dissipation theorem. The effect of space discretization into cells in numerical hydrodynamics on the spatial correlation of fluctuations is studied with the help of static structure factor. A difference from the continuum result is pointed out and systematically investigated. Time evolution of the ensuing fluctuations is examined and the Green-Kubo formula is used for the extraction of shear viscosity from the fluctuations. The extracted value is demonstrated to be the same as the input value of the shear viscosity for the simulations using Israel-Stewart equations. It is also shown that numerical regulators commonly used in the codes to ensure stability of the calculation severely spoil the fluctuations and break the fluctuation-dissipation theorem.