Information Theory
Papers
Fresh from arXiv, typeset properly. Highlighted words trade on the market.
All of arXiv
Monday 14 September
Let $Y_s=\sqrt{s}X+Z$, where $Z$ is standard Gaussian and independent of the real random variable $X$. We prove that, under the square-exponential moment condition $\mathbb{E}e^{\beta X^2}<\infty$ for some $\beta>0$, the scalar minimum mean-square error $\operatorname{mmse}_X(s)$ is analytic at zero signal-to-noise ratio if and only if $X$ is Gaussian, with constant random variables included as degenerate Gaussians. The proof converts estimation in the Gaussian channel into a backward heat flow acting on the moment-generating function $M(z)=\mathbb{E}e^{zX}$. Under the stated tail condition, every non-Gaussian input forces $M$ to have a nonzero complex zero. We show that each zero cluster produces a finite singularity in its localized Borel transform at the action $\xi=z_0^2/2$. After removing the action scale, the Borel coefficients have a nonzero $n^{-1/2}$ prefactor for a simple zero. A zero of multiplicity $m\geq 2$ splits according to the roots of a Hermite polynomial and instead contributes a prefactor $n^{-m/2}e^{r_m\sqrt{2n}}$. A finite-disc localization and relative-cycle continuation argument then show that at least one such singularity survives in the full Borel transform. Thus, for every non-Gaussian input in the stated class, the formal zero-SNR expansion is Gevrey-1 but divergent. Rational-MMSE rigidity and the analogous analyticity criterion for mutual information follow as corollaries.
Machine Learning
Structured Features Overfit Where Random Features Grok
Xu, Vardi and Safran (ICML 2026) prove that over-parameterized ridge regression over an unstructured random Gaussian feature map groks, with the delay between memorization and generalization growing as $1/\lambda$ in the weight decay. We show that on a structured feature map the same delay does not appear. For a band-limited Fourier feature map over $\mathbb{Z}_p^2$ carrying a single-character target that lies inside the expressible class, enlarging the band at fixed positive weight decay drives peak held-out accuracy monotonically from $1.00$ to $0.07$, with no memorize-then-generalize regime anywhere along the sweep. The degradation is not an interpolation effect. It sets in at capacity ratio $q/n = 0.638$, far below the interpolation threshold, on separate grounds from the exact null space that appears above it. What does have a sharp boundary is the active support. Holding the nominal dimension fixed and masking the band back to $1089$ active modes restores held-out accuracy of $1.000$ with zero variance across seeds, while the full $4225$-mode band collapses to $0.185$. The number of active modes acts through the teacher-weighted spectrum of the empirical Gram matrix and not through the capacity ratio, which makes this a statement about feature geometry and not a restatement of double descent.
Read the full abstract
Human-Computer Interaction
Personalizing Personal Health Interfaces: Co-Design with Generative AI
Personal health interfaces present wellbeing data through standardized dashboards that rarely fit how people interpret or act on it. Personalizing them to what people would like to see for themselves often requires design and technical expertise, a barrier that generative AI may potentially lower. Therefore, we ask what designs emerge and how it enables and constrains the design process. We conducted a co-design study where 14 participants redesigned Google and Apple Health interfaces using Figma Make. Participants reimagined interfaces that supported personal context, future planning, and interactive experiences, yet conversational AI designs converged around chat-window conventions. AI helped materialize loosely articulated ideas, but model defaults and generation latency shaped iteration. The process more readily operationalized interpretability and accountability than privacy, trust, and emotional safety. Generative co-design let participants create interfaces directly, blurring the boundary between intentions and model defaults. We discuss implications for preserving agency and flexible user-directed interfaces.
Read the full abstract
Computation and Language
Mirror, Mirror on the Wall: Prompt Echoing in Small Instruct Language Models
Prompt echoing is a recognized failure mode of instruct language models, in which a model instead of generating a response, mirrors the provided prompt, even though it did not receive a specific instruction to do so. Is this phenomenon a sign of the model leaking the content of its training dataset, or is it rather caused by a misaligned behavior of the internal induction/copying mechanisms? We investigate prompt echoing small language models from different families (Gemma, Llama, Qwen, SmolLM and OLMo) and show that echoing prompts are likely to have partial overlap with the training dataset but the phenomenon is primarily driven by the model's induction heads.
Read the full abstract
Machine Learning (Statistics)
Data Attribution via Sketched Metadifferentiation
Data attribution seeks to quantify how individual training examples shape a model's predictions and underpins problems including data valuation, machine unlearning, and model interpretability. Despite having a long line of work, computationally scalable methods often struggle to predict the effect of removing training data in neural networks due to their non-convex nature. To overcome this challenge, metagradient-based methods such as MAGIC (Ilyas and Engstrom, 2025) differentiate each prediction through the entire training run and compute its exact influence with respect to the training data, but require a separate run for every prediction. To reduce this cost, we cast budgeted attribution as estimating a large influence matrix from a small number of measurements. We show that the measurements most appropriate for recovering this matrix differ from those best suited for attribution itself. We then present two algorithms, MAGE and SPELL, suited for reconstruction and attribution respectively, that run on existing metagradient machinery at no extra cost. Empirical studies demonstrate strong performance over existing baselines across training scales and measurement budgets.
Read the full abstract
Plasma Physics
Phasor-particle-in-cell algorithm for bidirectional external-circuit coupling of inductively coupled plasma
We develop a self-consistent phasor-particle-in-cell algorithm to capture bidirectional inductive and capacitive feedback between an inductively coupled plasma and a distributed radio-frequency coil circuit. The method couples a two-dimensional axisymmetric particle-in-cell model with Monte Carlo collisions to a network that resolves the current and potential of each coil turn. Precomputed unit-current Helmholtz solutions provide the vacuum impedance matrix, while projection of the plasma-only field supplies the series back electromotive force. An established charge-based electrostatic coupling is extended to individual turns: surface charges obtained from a discrete Gauss law consistent with the Poisson solver yield shunt displacement currents through their fundamental harmonics. Both responses are accumulated over a complete radio-frequency period to update the relaxed circuit state once per period. Vacuum tests showed differences of -1.58 percent in single-solenoid inductance and -4.38 percent in mutual inductance relative to analytical references. A particle-free five-node manufactured benchmark with both coupling channels active yielded a scaled complex-state error of 1.90 x 10^-5 against the continuous reference and approximately second-order spatial convergence. Application to an unshielded argon reference cell reached statistically stationary states at coil-port powers of 60 and 100 W. Net capacitive heating accounted for 29.35 percent and 20.07 percent, respectively, of the combined inductive and capacitive plasma power. Independently accumulated field-side and circuit-side powers, including conductor loss, agreed within approximately 0.5 percent. These results support turn-resolved bidirectional field-circuit coupling for axisymmetric plasma simulations using a fundamental-frequency circuit representation.
Read the full abstract
Complex Variables
Rhaly operators between Dirichlet-type spaces: a complete classification
Let \(\eta=(\eta_n)_{n\ge0}\) be a complex sequence such that \(F_\eta(z)=\sum_{n\ge0}\eta_nz^n\in\Hol(\D)\), and let \(\mathcal R_{(\eta)}\) be the associated Rhaly operator. We give a complete characterization of boundedness and compactness of \(\mathcal R_{(\eta)}:\mathcal D^p_\alpha\to\mathcal D^q_\beta\) for \(1<p,q<\infty\) and \(\alpha,\beta>-1\). The characterization is formulated in terms of the \(H^q\)-norms of the dyadic blocks of \(F_\eta\). For \(-1<\alpha<p-2\), boundedness and compactness coincide and reduce to \(F_\eta\in\mathcal D^q_\beta\). For \(\alpha=p-2\) and for \(\alpha>p-2\), boundedness is characterized by membership in analytic truncated Besov spaces and analytic Besov spaces, respectively; compactness has the same characterization for \(q<p\) and is characterized by the corresponding little spaces for \(p\le q\). We also obtain operator norm and essential norm estimates. As applications, we obtain boundedness and compactness characterizations for Rhaly operators between weighted Bergman spaces and for Cesàro-type operators induced by positive measures. In particular, we characterize \(C_\mu:B^p\to B^p\), answering a question left open by Sun, Ye and Zhou and later noted by Tang. For each \(p>2\), we also construct a symbol \(F_\eta\in H^\infty\cap\lambda^p_{1/p}\) for which \(\mathcal R_{(\eta)}\) is not bounded on \(H^p\).
Read the full abstract
Astrophysics of Galaxies
The driving mode of turbulence in disc galaxy simulations with adaptive mesh refinement
Turbulence is a key ingredient in controlling the structure of the interstellar medium (ISM) and star formation, yet we still lack a detailed understanding of its drivers in galaxies. Previous idealised simulations of turbulence employ a stochastic forcing field to drive turbulence. The geometry of this forcing field - whether it is predominantly solenoidal or compressive - is a key parameter governing how turbulence shapes the ISM density distribution and regulates the star formation rate. The turbulence driving parameter ($b$) quantifies the relative contribution of compressive versus solenoidal driving.. Therefore, accurate knowledge of the driving parameter is essential for understanding and predicting star formation, and for sub-grid modelling of ISM physics and the star formation rate. In this work, we introduce an algorithm to measure the turbulence driving parameter in adaptive mesh refinement (AMR) simulations of galaxies. We focus our analysis on a synthetic Large Magellanic Cloud (LMC), present-day analogue with a total mass $M \approx 10^{11}\,\mathrm{M}_\odot$. We find that turbulence is driven primarily solenoidally ($b<0.4$) within the inner $\sim4\,\mathrm{kpc}$ of the galaxy, and becomes increasingly compressive with $b>0.4$ towards the outskirts, $R\gtrsim4.5\,\mathrm{kpc}$. The volume-weighted median across the disc, $b\simeq0.4$, is consistent with the natural mixture of driving modes. We further find that $b$ is weakly correlated with the strength of shear, in that solenoidal driving tends to be associated with regions of higher shear, as expected for the more central parts of galaxies. These trends persist over $\sim\!2\,\mathrm{Gyr}$ of the galaxy's evolution.
Read the full abstract
Quantum Physics
Scaling up multi-mode entanglement generated by mode swapping
Entanglement between two local multi-photon multi-mode systems can be generated by swapping a pair of modes between the two local multi-mode systems. In this presentation, we consider possible strategies for the efficient generation of entanglement between systems with three or more modes. It is shown that the choice of photon number inputs in the local systems adds a new degree of freedom to the non-local interference effects observed in the output photon number statistics.
Read the full abstract
Cryptography and Security
SpliTEE: Fast and Private LLM Inference by Coupling GPU-Assisted Trusted Execution Environments with Differential Privacy
User prompts provided to large language models (LLMs) may contain private information. One way to protect them is to execute the LLM inside a trusted execution environment (TEE). However, this results in slow inference times as current TEEs are significantly slower than GPUs for LLM inference. To circumvent this, Tramèr and Boneh (2019) proposed Slalom which splits neural network inference between a TEE and an untrusted GPU. They encrypt inputs to computations outsourced to the GPU. In this paper, we extend this split-inference architecture to LLM inference and instead protect intermediate inputs using differential privacy (DP). We first demonstrate that masking intermediate representations is necessary by showing an 80% accuracy on a prompt-reconstruction attack from these representations. Our main contribution is a global sensitivity analysis of key functions in LLM inference, which bounds the required scale of DP noise. Unlike encryption, DP avoids quantization, allowing the LLM to remain in the floating-point domain. We also derive an upper bound on the floating-point error from masking and subsequent noise cancellation as a function of the privacy parameter epsilon, keeping the same quality of the LLM response. We implement our architecture using the Intel TDX TEE and two LLMs: Llama-3.2-3B and Qwen3-4B. Our split execution is nearly twice as fast as fully TDX-based inference. Moreover, it is at most 43% faster than Slalom while achieving higher accuracy. Finally, we demonstrate that prompt reconstruction, even with knowledge of the DP mechanism, cannot recover more information than is contained in an unrelated prompt.
Read the full abstract
Artificial Intelligence
The average-farmer illusion in language-model simulations of agricultural decisions
Language-model agents are increasingly used as synthetic people in surveys and social simulations, yet their apparent realism is often judged from population averages or distributional similarity. We tested what such evidence actually establishes by comparing Claude, Codex and Kimi under four prespecified prompt designs with matched farmer decisions from China and four African countries. Some configurations reproduced observed means and adoption rates. However, their person-level predictions were weak; their decisions clustered around typical values and policy-relevant extremes were largely missing. Most strikingly, a simple generator fitted only to the observed marginal dis- tribution, and given no information about any farmer, achieved greater distributional similarity than every language-model configuration. Prompt additions produced conditional gains rather than uni- versal improvement: results varied with model, outcome, population and validation target. We call this the average-farmer illusion: a synthetic population can look realistic while failing to repro- duce who does what or how behaviour varies. We provide a claim-matched validation framework and reusable modular prompts that turn prompt construction into an auditable experimental process. Population-level resemblance should therefore be treated as the start of validation, not as evidence of individual simulation.
Read the full abstract
Machine Learning
MoARa: Module-Aware Rank Allocation and Structure-Preserving Decomposition for Low-Rank LLM Pre-training
Low-rank gradient projection reduces the optimizer-state memory cost of large language model (LLM) pretraining, but the steps and wall-clock time needed to reach a target quality remain a meaningful axis for improvement. We attribute this to two design choices in existing methods: the projection-rank budget is allocated uniformly across Transformer modules with heterogeneous projection sensitivity, and projecting a raw gradient attenuates its magnitude and direction jointly. We propose MoARa, which combines a static profiling-based module-aware projection-rank allocation with a block-wise magnitude-direction decomposition; the default block size is set in the neighborhood of the attention head dimension. Across five Transformer architectures spanning Llama, Qwen, and DeepSeek at 300M to 7B scales, GaLore with MoARa reaches standard GaLore's final perplexity in 37% fewer steps and 34% less wall-clock time on Llama 2 7B, with only 0.2% peak reserved memory overhead under standard graph compilation. Across the six low-rank pretraining methods we evaluate, module-aware rank allocation alone delivers directionally consistent step reductions on all six. On compatible hosts, the two-component design reaches up to 41.7% step reduction and 37.1% wall-clock reduction.
Read the full abstract
Materials Science
Thermally stimulated depolarization in hafnium oxide: experiment and numerical simulation
This paper presents an experimental and theoretical study of thermally stimulated depolarization (TSD) in hafnium oxide. Three different models for the probability of trap ionization are considered: the Frenkel effect and two variants of the multiphonon trap ionization mechanism, the multiphonon ionization mechanism of isolated traps at low trap concentrations and phonon-assisted tunneling between neighboring traps at high trap concentrations. The best agreement between the TSD experiment and theory is observed for the multiphonon ionization mechanism of isolated traps. Similar thermal energies of electron and hole traps in HfO2 are obtained ~ 1.3 eV.
Read the full abstract
Artificial Intelligence
Horizon-specific Expert Fusion for Photovoltaic Power Forecasting
Short-term photovoltaic power forecasting requires models to represent regular solar cycles and weather-driven fluctuations whose importance changes with the forecast horizon. This study develops a hierarchical ensemble that combines temporal neural models, historical analogs, state climatology, and gradient-boosted trees. Solar geometry and numerical weather forecasts describe the expected generation conditions, while horizon-specific convex weights combine complementary predictions. A separate calibration step uses available historical forecast errors to account for recent bias. The framework is evaluated on public PVDAQ data at 15–240-minute horizons and on three GEFCom2014 solar zones at hourly horizons up to four hours. On PVDAQ, the ensemble achieves a daylight capacity-normalized mean absolute error of 4.315%, reducing error by 4.11% relative to full-feature LightGBM and by 6.03% relative to fine-tuned Chronos-2 under identical calibration. Expert-removal experiments identify redundancy within the ensemble. Across three training seeds on GEFCom2014, learned fusion improves upon equal weighting but performs comparably to LightGBM. The results support horizon-specific combination as a useful forecasting strategy while showing that its advantage over strong individual models depends on the dataset and evaluation period.
Read the full abstract
Databases
FastPair: GPU-Optimized String Decoding
Modern data systems compress data at rest and decompress it only when needed to preserve interconnect bandwidth. This design is often inefficient on GPU-based compute platforms because many conventional compression techniques exhibit serial data dependencies that limit GPU parallelism, leaving resources idle. Recent NVIDIA GPUs address this decoding deficiency through the Decompression Engine (DE), an on-die, fixed-function decompression accelerator for general-purpose compression formats such as Deflate, LZ4, and Snappy. Recent work has proposed string codecs that replace frequent substrings with fixed-width codes from a small, trained dictionary, making each code's lookup independent. While these lookups can run in parallel, the resulting scattered reads and short output writes still do not align well with GPU hardware, which handles contiguous memory accesses more efficiently. We present FastPair, a GPU decoder that optimizes the existing dictionary decoding process by reorganizing lookups and assembling decoded substrings for contiguous output writes. On a B300, FastPair decodes ten real-world columns 2.4 to 4.2x faster than the DE, reaching up to 1.6 TB/s.
Read the full abstract
Atomic Physics
Phase-Sensitive Heterodyne Detection of MW using EIT Harmonics in Rydberg Atoms
We investigate the generation and characterization of higher-order harmonics in the probe-laser absorption arising from nonlinear interactions in an electromagnetically induced transparency (EIT) ladder system involving Rydberg states and driven by two microwave fields in the heterodyne configuration. We characterize the amplitude and phase of the generated harmonics as a function of the relative frequency and phase of the applied microwave fields. The phase of the $n^{\mathrm{th}}$ harmonic follows the relation $\phi_n=n\phi$, demonstrating phase multiplication and suggesting that higher-order harmonics may offer an enhanced phase response for phase-sensitive measurements. We further characterize the harmonic amplitudes and bandwidths and find that the measured bandwidths are significantly larger than the intrinsic Rydberg-state linewidth, consistent with power broadening under the experimental conditions. The experimental observations are supported by density-matrix calculations, which reproduce the key features of the measured harmonic response.
Read the full abstract
Computer Vision
Tele360: Real-Time Feed-Forward Human Reconstruction from Sparse Unposed Cameras
Live free-viewpoint visualization of real humans is critical for immersive communication and interactive digital experiences. Existing methods either rely on computationally expensive optimization or require calibrated cameras and low-resolution inputs, making real-time high-resolution deployment impractical. In this work, we present Tele360, the first real-time feed-forward system for dynamic human reconstruction and live free-viewpoint visualization from sparse, unposed RGB streams. Our system jointly estimates camera poses and reconstructs a dynamic 3D Gaussian representation for each time instance in a single forward pass. To achieve this, we start by designing a lightweight sparsity-aware multi-view transformer backbone that tokenizes foreground human regions while preserving global context through a shared scene token. We then employ a fully transformer-based Gaussian decoder to mitigate convolution-induced over-smoothing while keeping decoding sparse and efficient. In addition, we introduce a hybrid feature pyramid that injects multi-scale appearance cues into geometry prediction. We further introduce a lightweight differentiable Levenberg-Marquardt camera refinement layer to enhance multi-view consistency and geometric alignment. Moreover, to stabilize learning under sparse, unposed inputs, we transfer multi-view geometry priors from a large visual-geometry foundation model via teacher-student distillation. Finally, the predicted Gaussian maps are streamed with video codecs to remote devices for interactive free-viewpoint rendering. Extensive experiments show that Tele360 achieves state-of-the-art visual quality on studio benchmarks while supporting real-time 2K input-to-rendering at over 25 FPS on a single consumer GPU. Additional captured sequences illustrate its performance across varied subjects, clothing, and motions under our multi-camera setup.
Read the full abstract
High Energy Physics (Experiment)
Search for charmonium(like) states $X$ in $e^{+}e^{-}\rightarrow\gamma X\rightarrow\gamma D^{*0}\bar{D}^{*0}$ at BESIII
A search is performed for a state $X$ decaying into $D^{*0}\bar{D}^{*0}$ produced in the process $e^{+}e^{-}\rightarrow\gamma X$ using a data sample corresponding to an integrated luminosity of 1667.4 $\rm pb^{-1}$ collected at $\sqrt{s} = 4.682$ GeV with the BESIII detector at the BEPCII. The state $X$ could be one of the $C$-even states $X(4013)$, $\eta_{c}(3S)$, $\chi_{c0}(3P)$, $\chi_{c1}(3P)$, or $\chi_{c2}(3P)$. No significant signal is observed in the corresponding signal region. Upper limits of $\sigma_{e^{+}e^{-}\rightarrow\gamma X}\cdot {\rm Br}_{X\rightarrow D^{*0}\bar{D}^{*0}}$ at 90% confidence level are provided, where $\sigma_{e^{+}e^{-}\rightarrow\gamma X}$ represents the cross section of the $e^{+}e^{-}\rightarrow\gamma X$ process, and ${\rm Br}_{X\rightarrow D^{*0}\bar{D}^{*0}}$ is the branching fraction of the $X\rightarrow D^{*0}\bar{D}^{*0}$ process.
Read the full abstract
Distributed Computing
Validating Hybrid-State Cache Recovery for GLM-5.3-Flash with vLLM and LMCache
External cache transfers can succeed while a hybrid language model resumes from an inconsistent state. We examine the full 45-layer GLM-5.3-Flash model, using the RedHatAI/ GLM-5.3-Flash-NVFP4 quantized checkpoint with vLLM and LMCache under four-way tensor parallelism. A complete-hit recovery mismatch restored state for the full prompt while the scheduler credited one fewer token. We aligned recovery through strict-prefix lookup and established a numerical comparison using shared computation corrections, matched checkpoint scheduling, and fixed per-rank kernel configurations. In a nine-length serial workload, agreement with the modified recomputation control improved from 34/36 to 36/36 generations, each containing 64 token IDs. A separate instrumented run passed recorded transfer-page, effective-tail, and delayed-save checks. Three additional synthetic templates passed 72 paired 256-token continuations across two fresh-container runs. A subsequent serial performance study preserved output equality across 120 requests; among the measured trials, CPU reload reduced time to first token by 46-64% and total request time by 1.9-7.0% relative to modified cold recomputation. The contribution is an experimentally validated integration repair applying an existing checkpoint-alignment principle. The evidence is confined to one model revision and controlled configuration; it does not establish general determinism, task-quality equivalence, concurrent-serving gains, or capacity beyond GPU memory.
Read the full abstract
Machine Learning
Pick Your Poison: Learning to Select Poison Sets for Stronger LLM Backdoor Attacks
Backdoor poisoning attacks add poisoned examples to otherwise-clean finetuning data, pairing a trigger with a target behavior that the model learns to produce when the trigger appears. Existing evaluations typically fix the number of poisoned examples and sample them at random from a candidate pool. We show that this can severely underestimate worst-case vulnerability: across three LLaMA-3-8B backdoor settings, holding the model, clean data, and poison count fixed, attack success ranges from 3% to 80% depending only on which poison set is chosen. We formalize poison selection as oracle-budgeted set optimization and introduce SAILS (Set-level Audit-Informed Iterative Learned Selection), which learns a set scorer from a few hundred finetune-and-evaluate runs, ranks millions of candidate sets, and audits only a small shortlist. SAILS improves held-out attack success by 30 percentage points on average over the strongest influence baselines, transfers from small-scale to full-scale finetuning, and extends to code-generation, agentic, and API-only backdoors.
Read the full abstract
Computer Vision
TwinICL: Diagnosing Multimodal In-Context Learning through Paired Counterfactuals
In-context learning (ICL) enables models to infer tasks from demonstrations, but existing benchmarks generally lack matched text and image versions needed to compare ICL performance across modalities. We introduce TwinICL, a procedurally generated benchmark providing such pairs for controlled comparison. Across six open-weight models and 38 tasks, multimodal ICL consistently underperforms text-only ICL, with gaps varying by task family. To test whether this gap can be recovered, we target visual access, task framing, and reasoning through three interventions. Their combination recovers strong multimodal ICL performance on a diagnostic subset, despite limited or inconsistent individual effects. To distinguish difficulties in executing tasks from those in inferring them, we evaluate models with explicit task instructions, revealing a modality gap even when the task is known. We then examine how adding demonstration inputs and outputs reshapes this gap, highlighting demonstrations' dual role as additional context to process and evidence about the task. The dataset is available at https://github.com/lab-flair/TwinICL.
Read the full abstract
High Energy Physics (Phenomenology)
Inelastic Dark Matter at LZ from Radiative Dirac Neutrino Mass Paradigm
The LUX-ZEPLIN (LZ) experiment has reported a high-energy nuclear recoil candidate, LZ230616, at $E_R = 248 \pm 23_{\rm stat} \pm 23_{\rm sys}\,$keV, which is difficult to reconcile with elastic scattering of halo dark matter (DM). Endothermic inelastic scattering offers a natural explanation, but it rests on a nucleon coupling that is off-diagonal rather than diagonal. We show that this feature arises automatically within a radiative Dirac neutrino mass framework. The Standard Model is extended by three right-handed neutrinos (RHNs), a pair of vector-like neutral fermions for each generations, and a scalar sector comprising an inert doublet and a real singlet, governed by a $\mathcal{Z}_2 \times \mathcal{Z}_4$ symmetry. The $\mathcal{Z}_2$ symmetry stabilizes the DM, while the $\mathcal{Z}_4$ symmetry forbids the tree-level Dirac Yukawa coupling together with all renormalizable Majorana mass terms, thereby allowing the generation of Dirac neutrino masses at the one-loop level through a softly broken scalar trilinear coupling `$\kappa$'. Since the neutral dark-sector fields can be expressed in terms of real scalar mass eigenstates, the $Z$ boson couples to them purely off-diagonally and elastic $Z$-mediated scattering is absent identically rather than simply suppressed. The singlet–doublet mixing induced by `$\kappa$' governs both the inelastic rate and the one-loop Dirac neutrino mass, which therefore vanish as $\kappa \to 0$. We find that DM masses in the few hundred GeV to TeV range, with splittings of $\mathcal{O}(340-360)\,$ keV, simultaneously reproduce the observed relic abundance, account for the LZ event and yield neutrino masses of the correct order, while respecting elastic direct-detection limits and all relevant theoretical and experimental constraints. A substantial part of the surviving parameter space lies within reach of DARWIN.
Read the full abstract
High Energy Physics (Phenomenology)
Vacuum Polarization Effects from Dark Fermion Loops in $Z'$ Mediator Models and their Impact on Angular Observables in $B \to K^* \mu^+\mu^-$ Decays
Rare flavour changing neutral current decays such as $B \to K^{*}\mu^{+}\mu^{-}$ provide sensitive probes of physics beyond the Standard Model (SM), being forbidden at tree level and arising only through loop induced processes. Angular analyses yield observables sensitive to short distance dynamics. The optimized observable $P'_5$ shows a persistent $\sim 3\sigma$ deviation from SM predictions in $q^{2}\in[1.1,6.0]\,\mathrm{GeV}^{2}$, motivating modified Wilson coefficients. A theoretically motivated framework is investigated in which vacuum polarization of a Dirac dark fermion $\chi$ modifies the propagator of a heavy $Z'$ boson, inducing a momentum dependent contribution to $C_{9}(q^{2})$. The loop corrected propagator is incorporated into the effective weak Hamiltonian for $b\to s\ell^{+}\ell^{-}$. Bin averaged angular observables are computed with flavio, and $34{,}997$ parameter points are scanned over different chiral structures of the $Z'$ couplings. A $\chi^{2}$ analysis identifies preferred regions that are vector dominated and chirally asymmetric. The best fit solution significantly alleviates the $P'_5$ tension, reducing the deviation in the central $q^{2}$ bin to $0.12\sigma$, an improvement of approximately $96\%$ relative to the SM. This improvement is driven by tree level $Z'$ exchange, which induces $\Delta C_{9}\simeq-1.2$, consistent with global fits and compatible with other angular observables and $B_{s}\to\mu^{+}\mu^{-}$. Dark fermion vacuum polarization adds a subleading $q^{2}$ dependence to $C_{9}$; it does not dominate the fit, but supplies a distinctive threshold structure testable with finer binning. These results show that a $Z'$ portal to a dark fermion provides a phenomenologically consistent description of the $P'_5$ anomaly and a concrete benchmark for LHCb Run~3 analyses.
Read the full abstract
Data Structures and Algorithms
Matrix Spencer: Eight Standard Deviations Suffice and an Almost-Linear Time Algorithm for Dense Input
The Matrix Spencer conjecture asserts that for all symmetric matrices $A_1,\ldots,A_n\in\mathbb{R}^{n\times n}$ with $\|A_i\|\le1$ there are signs $\varepsilon_1,\ldots,\varepsilon_n\in\{-1,1\}$ with $\|\sum_{i=1}^n\varepsilon_iA_i\|=O(\sqrt n)$. We prove it: a signing of discrepancy below $8\sqrt n$ always exists. We also give a randomized algorithm that finds a signing of discrepancy below $12\sqrt n$ with failure probability at most $p$. The algorithm uses $n^{3+o(1)}\operatorname{polylog}(1/p)$ arithmetic operations in the real-arithmetic model. This matches the size $n^3$ of the dense input up to subpolynomial factors. In the other direction, we prove that for every $n$ there are collections of symmetric matrices such that every signing has discrepancy at least $(2-o(1))\sqrt{n}$. We present three different proofs of the matrix Spencer conjecture. The key to every proof is a hereditary small-ball estimate. This is a lower bound on the Gaussian measure of the spectral body $\{x\in \mathbb{R}^n:\|\sum_ix_iA_i\|\le R\}$ that holds for every subfamily of the matrices. The other ingredient turns that Gaussian measure into a partial signing. We give three approaches to obtain such a signing. The first one covers the cube by partially signed faces through Gaussian concentration with a constant $7\cdot10^9$. The second proof replaces the covering by a projection lemma with explicit parameters for a constant $156000$. The third proof turns Gaussian measure into signs by a lossless coding, with no union bound. It proves the estimate at the right radius with smooth spectral barriers and certified coefficients. It gives a constant below $7.88$. For algorithms, the main idea is to project Gaussian points onto a smoothed spectral body. The $n^{3+o(1)}$ time algorithm tracks the Gibbs matrix of that body across coordinate-descent steps with sketched increments and random refreshes.
Read the full abstract
Number Theory
Ekedahl-Oort strata meeting the supersingular locus
We give an explicit combinatorial criterion for an Ekedahl-Oort (EO) stratum in the moduli stack of principally polarized abelian varieties of dimension $g$ in characteristic $p > 2$ to meet the supersingular locus. The criterion is obtained by translating a non-emptiness criterion for basic affine Deligne-Lusztig varieties, into an explicit condition on the elementary sequences indexing the EO strata. Furthermore, we give explicit bounds for the number of EO strata of $p$-rank zero which are disjoint from the supersingular locus. As a result, we prove that asymptotically almost every EO stratum of $p$-rank zero meets the supersingular locus. As an application, we also give a lower bound for the number of EO strata meeting the supersingular locus on unitary PEL Shimura varieties of signature $(a,b)$.
Read the full abstract
Probability
From torpid to rapid mixing: group averaging for a weakly interacting Ising star
We study lazy single-site Metropolis dynamics $P_{\beta}$ at inverse temperature $\beta \geq 0$ on $\{-1,+1\}^d$ for an Ising star with additional signed interactions among the leaves. If the absolute row sums of the leaf-interaction matrix are at most $\kappa\le1/2$, the worst-case total-variation mixing time of $P_{\beta}$ is at least of order $d\exp\{c\beta(d-1)\}$ for $\beta\ge1$, with universal $c>0$. Averaging over global spin reversal reduces the mixing time to $\mathcal O(d^2(1+\beta))$ for both $GP_\beta G$ and $(P_\beta+G)/2$, where $G$ is the Gibbs kernel induced by the partition of the state space into spin-reversal orbits $\{-x,x\}$. Partition-function interpolation gives the lower bound for $P_\beta$. The upper bounds follow from Wu's Dobrushin inequality and a decomposition into projection and restriction chains. This gives an explicit example in which group averaging turns an exponentially slow chain into a polynomially fast one.
Read the full abstract
Computation and Language
SALUTE: Benchmarking and Adapting LLMs for the Defense Domain
Defense is a knowledge-intensive domain that requires precise understanding of specialized terminology, doctrinal concepts, operational procedures, and evolving military events. Although recent work has explored language technologies for military applications, existing efforts remain fragmented: they are often task-specific, rely on limited adaptation pipelines, or lack comprehensive defense-domain evaluation. In this paper, we present SALUTE, an end-to-end framework for benchmarking and adapting LLMs for the defense domain. SALUTE integrates Salute-Corpus, a curated corpus from open-access U.S. military doctrine and government documents; Salute-Conv, a grounded instruction dataset from doctrinal sources and decade-long defense news; Salute-Pref, a defense-aware preference dataset; and Salute-Bench, a rigorously filtered benchmark for evaluating defense-domain understanding and reasoning over doctrine and defense news. Based on these resources, we train Salute-LLM through multi-stage post-training with continual pretraining, supervised fine-tuning, and preference alignment. Extensive experiments show that Salute-LLM achieves strong defense-domain performance while retaining competitive general capabilities, demonstrating the effectiveness of SALUTE as an end-to-end framework for defense-domain LLM adaptation.
Read the full abstract
Distributed Computing
Shared KV Caching for Replicated 27B Inference: Correctness Failures and Performance Boundaries
Shared host-memory caching can avoid repeated prefill when a request moves between inference replicas. Its usefulness depends on both correct state transfer and lost prefix locality. We study two single-GPU 27B vLLM replicas sharing a 256 GiB LMCache pool. After adopting an existing packed-page patch, we isolate a raw-pointer fallback that omits the dependency on the current CUDA stream. Controlled byte tests fail under an imposed delay and pass when the dependency is restored; the existing mixed allocator provides a working deployment path. Full-pool allocation checks and service regression complete the validation. A four-block OFF-ON-ON-OFF comparison contains 768 measured requests within two block pairs. Median cross-replica time to first content token falls from 31.715 to 0.605 seconds at 128k input and from 92.047 to 0.790 seconds at 256k. Six-turn synthetic sessions alternating replicas improve by approximately 35% and 45% at initial contexts of 32k and 128k, while fixed placement shows little benefit. This engineering case study identifies practical validation steps and the locality conditions in which shared caching pays off.
Read the full abstract
Cryptography and Security
Efficient Branch-and-Bound Testing and Verification of zkVMs
Zero-knowledge virtual machines (zkVMs) enable verifiable execution of general-purpose programs by translating virtual machine semantics into algebraic constraints over execution traces. The correctness of these constraints is critical: a single incorrect constraint can admit forged proofs (under-constrained) or reject valid executions (over-constrained). Existing approaches do not provide meaningful guarantees at production scale: fuzzers and unit tests often miss bugs, SMT solvers struggle with the size and nonlinearity of constraints, and theorem provers require substantial manual effort. We present ZEBRA, a fully automated verification and bug-detection framework: for a given program and input, the constraint must admit exactly one valid execution trace - no more and no fewer. This reduces zkVM verification to a solution-set cardinality problem over a canonical trace space, where redundancies such as null-row padding and non-deterministic permutations are eliminated prior to counting. To compute cardinality tractably, ZEBRA lifts analysis from finite-field witnesses to an integer interval lattice, exploiting a structural sparsity property of zkVM constraints: across 5 real-world zkVMs, constraints utilize only 14.0% of their theoretical connectivity capacity on average. This sparsity enables tight interval propagation with limited approximation error. ZEBRA performs a parallel branch-and-bound search that either produces a concrete counter-example or certifies the absence of violations within a bounded region. We evaluate ZEBRA on five real-world zkVMs. ZEBRA discovers 11 zero-day bugs; 6 have already been independently confirmed and 3 have been fixed by developers. Compared to SMT-based verification, ZEBRA is 51.5x faster, verifies 16.5 percentage point more instances, and its range verification provides up to 63x efficiency gain over repeated single-input verification.
Read the full abstract
Theoretical Economics
Contracting for Information: Heterogeneous Costs and Investment Opportunities
A principal faces a decision problem under uncertainty and can contract with a researcher to provide relevant information. The cost of acquiring information is only known to the researcher, and, moreover, by privately making an investment the researcher can reduce their expected cost. We characterize optimal contracts by reducing the problem to information design with cost constraints. The principal induces investment below a cutoff in its sunk cost. With two cost types, investment may be underprovided but is never excessive relative to the first best. With more types, the cutoff property survives but overinvestment can occur. Investment incentives also reshape information acquisition: optimal experiments may need to generate more distinct beliefs than there are possible states, and their design can be sensitive to the principal's initial beliefs.