Peer Revenue Publish or profit

Papers

Fresh from arXiv, typeset properly. Highlighted words trade on the market.

All of arXiv

Friday 4 September

High Energy Physics (Phenomenology)

Exothermic and Endothermic Inelastic Dark Matter Interpretations at LZ: Sideband Constraints and Future Prospects

The LZ experiment has extended its nuclear-recoil search to 270 keV and reports a single nuclear-recoil-like event, LZ230616, at $\sim250$ keV, in a region with negligible expected background. We show that exothermic inelastic dark matter, in which the ambient dark state down-scatters and releases its mass splitting $|\delta|$ as recoil energy, can naturally produce a recoil peak at this energy while remaining consistent with the null result at lower energies. Our simple fit leaves the mass unconstrained, because a heavy candidate can hide its peak above the LZ region-of-interest. However, we show that the empty high-energy sideband, which LZ uses for background validation, disfavors broadly peaked spectra. For an exothermic explanation, the sharper peaks produced by lighter dark matter with larger mass splittings are therefore preferred. Realizing this scenario requires cosmologically stable states and a leptophobic mediator. Normalizing to the one observed event, we predict $\simeq4.5$ signal events in LZ's projected 1000 live-day exposure, and show that existing XENONnT and PandaX-4T data could already test the interpretation in an extended analysis window.
Read the full abstract

Machine Learning

WEECFP-SuRGE: A Position-Aware Substructure Encoding Method for Molecular Property Prediction

Computational molecular property prediction requires representations that capture local chemistry, long-range interactions, and molecular topology. Conventional fingerprints provide efficient local substructure features, whereas learned graph and sequence models can represent broader context but often rely on pretraining or three-dimensional conformers. We introduce Wide Encoded Extended Connectivity Fingerprints (WEECFP) with Substructure Rotary Graph-distance Encoding (SuRGE), a tokenized hierarchical Morgan representation in which graph-distance-dependent rotations are applied at the input and within transformer self-attention. Across MoleculeNet and the Therapeutic Data Commons ADMET benchmarks, WEECFP-SuRGE is competitive with recent pretrained and geometry-aware methods without external pretraining or conformer generation. We also show that the rotated token representation remains structurally informative. A guided confirmed-handshake overlap procedure reconstructs the correct constitutional isomer for 92.6% of a 4,200-molecule self-library evaluation. Together, the predictive and reconstruction results indicate that WEECFP-SuRGE preserves local substructure identity while making relative topology available to the model.
Read the full abstract

Combinatorics

Curvature Diffusion of Inverse-weight Lin–Lu–Yau Ricci Flow on Finite Trees

We study the inverse-weight Lin–Lu–Yau Ricci flow on finite trees, and prove the convergence of the curvature on every edge via its diffusion equation. Moreover, the limiting curvature is the unique minimum-norm point of a polyhedron determined by the tree. We characterize possible limits of the normalized edge weights, give criteria for their convergence, and show that different positive initial metrics have equivalent omega-limit sets.
Read the full abstract

Algebraic Geometry

Cohomology on Cotangent Bundles of Partial Flag Varieties in Type A

Let $G=\mathrm{SL}_n(\mathbf C)$ and let $P\subset G$ be a standard parabolic subgroup with Levi factor $L$. For a $G$-dominant weight $\lambda$, consider the vector bundle on $T^*(G/P)$ obtained by pulling back the vector bundle on $G/P$ associated to the irreducible $L$-module $V_L(\lambda)^*$. We express its cohomology as a direct limit of the cohomology of certain line bundles on a Bott–Samelson variety associated to an affine Kac-Moody group. This comparison yields vanishing of higher cohomology and shows that the global sections are generated over $\mathbf C[\mathfrak g^*]$ by their degree-zero part $V_G(\lambda)^*$. Using these results together with the Braverman–Kazhdan intertwiners constructed in earlier joint work with A. Slipper, we give an explicit generating set for $\mathbf C[T^*(\mathrm{SL}_n/[P,P])]$. Finally, in an appendix joint with Tom Gannon, we combine these results to show the affinization $\mathrm{Spec}(\mathbf{C}[T^*(SL_n/[P,P])])$ has terminal singularities.
Read the full abstract

Information Theory

Ultra-High Resolution Method for Multipath Within a Co-Delay-Doppler Bin in DFT-P-OCDM

Communication systems can reuse their transmitted signals for sensing without dedicated radar transmissions. For an established DFT-preprocessed orthogonal chirp division multiplexing (DFT-P-OCDM) waveform, this task becomes difficult when several physical paths in a doubly selective channel fall into the same co-delay-Doppler bin. In this case, the number of resolvable delay classes inferred from the pilot may be smaller than the number of physical paths within the co-bin. This mismatch increases the difficulty of path number determination, fractional Doppler offset estimation, and channel reconstruction. We derive a pointwise relationship between the input and output in the DFT preprocessed Fresnel (DPF) domain for doubly selective channels with multiple paths within the co-delay-Doppler bin. Based on this input and output relation, we propose the two stage ultra high resolution (TSUR) framework. The first stage uses pilots of the phase progression to estimate delay, while the second stage uses the leakage samples to estimate the Doppler and the number of paths within each delay class. Furthermore, we derive CRLBs and analyze how the pilot configuration trades sensing resolution and communication recovery. Simulation results demonstrate that TSUR resolves same delay paths within a co-delay-Doppler bin, remains robust to exist delay offsets, and achieves lower fractional Doppler estimation errors than sequential extraction and off-grid baselines.
Read the full abstract

Statistical Mechanics

Breakdown of the Overdamped Approximation in Fluctuating Environments

The overdamped approximation is widely used to describe Brownian motion in complex environments, but its validity becomes nontrivial when the friction coefficient itself fluctuates in time. We investigate this problem by comparing underdamped and overdamped Langevin dynamics subject to the same fluctuating friction and satisfying the fluctuation–dissipation relation at a common temperature. We show that environmental averaging and inertial elimination generally lead to different long-time transport when environmental fluctuations are fast compared with velocity relaxation. For rapidly fluctuating friction, the underdamped dynamics is governed by the arithmetic mean friction and yields $D_{\rm eff}^{\rm under}=k_{\rm B}T/\langle\gamma\rangle$, whereas the overdamped dynamics gives $D_{\rm eff}^{\rm over}=k_{\rm B}T\langle\gamma^{-1}\rangle$. For a two-state Markov friction, we derive the finite-time effective diffusion coefficient exactly and identify the full crossover between these two regimes, controlled by the competition between velocity relaxation and environmental switching. We further establish the fast-fluctuation result for general stationary friction processes and verify it for a continuous log-Ornstein–Uhlenbeck environment, for which the characteristic crossover time can also be determined independently. Our results reveal a noncommutativity between environmental averaging and inertial elimination, establish a timescale-dependent criterion for the validity of overdamped dynamics in temporally heterogeneous environments, and show that rapid environmental fluctuations can enhance, rather than suppress, the consequences of inertia.
Read the full abstract

Artificial Intelligence

ERPBench: Evaluating LLM Agents for Enterprise Decision-Making Across Competitive Market Ecologies

Large language model (LLM) agents are increasingly proposed for enterprise workflows, yet existing evaluations rarely test whether business-decision conclusions transfer across competitive market ecologies. We introduce ERPBench, an execution-instrumented benchmark for enterprise decision agents in a six-round Enterprise Resource Planning (ERP) simulation with coupled pricing, production, procurement, inventory, finance, and shared-market competition. ERPBench evaluates the same 100 fixed problems in two matched competitive market ecologies: Solo, where each evaluated LLM agent competes against fixed rule-based opponents, and Arena, where six evaluated LLM agents compete in a shared market. Across six model families, this yields 1,200 model-level trajectories spanning 7,200 decision rounds. Under the observed service configuration, the leading model differs between ecologies: DeepSeek leads in Solo (252.29M mean valuation; mean rank 1.67), whereas Gemini leads in Arena (263.95M; 1.76). The two ecologies identify the same task-level winner on only 21 of 100 problems, and Gemini's bottom-rank rate falls from 22 % to 0 % in Arena. ERPBench supports paired evaluation of whether enterprise-agent rankings transfer across competitive market ecologies, supplemented by aggregate execution-intervention analysis. Code and benchmark resources are available in our https://github.com/GAIR-NLP/erp-bench.
Read the full abstract

All authors Xinran Zhang, Pengrui Lu, Lyumanshan Ye, Pengfei Liu

Optics

Research and simulation of analytical polarization control enabled by optical computing on an integrated photonics chip

Dynamic polarization controllers are key devices with broad applications in many fields. However, most on-chip polarization controllers still rely on traditional blind-search methods, whereas analytical optical-computing approaches remain insufficiently explored, particularly with respect to calibration and endless polarization control. With the accurate relative phase of Mach-Zehnder interferometer (MZI) being fully controllable on an integrated photonics chip, we present an analytical polarization control (APC) method using four phase shifters and optical computing, eliminating the need for the traditional inefficient blind-search procedure. The basic structures and operations of APC are clarified. The proposed calibration method and endless control method enable continuous APC while compensating for phase differences within the MZI structures. We simulate the influence of the endless control unit on polarization control and quantify the effect of the fourth phase difference on the output extinction ratio. With the fourth phase shifter, the phase difference encountered during Stokes vector measurement can be effectively compensated, and rotations around all three axes on the Poincaré sphere can be realized. These results establish a practical APC architecture based on optical computing for photonics chips. The proposed APC methods, combined with a FPGA-based hardware acceleration, will enable high speed on-chip polarization controllers.
Read the full abstract

All authors Xueying Ren, Meinan Guo, Xuyang Wang, Bailin Shen, Lingyan Zhang, Minyue Yang, Nannan Ning, Jiaxin Huang, Lv Lv, Jun Zou, Yongmin Li

Artificial Intelligence

Harness-agnostic detection and immunization of reward hacking in self-evolving language models

Self-evolving language models improve by proposing candidate updates and keeping whatever raises a visible score. When that score is an imperfect proxy for the capability one actually wants, sustained selection widens the gap between the two. This is reward hacking. We introduce HackProbe, a monitor that attaches to an arbitrary self-evolving loop through two black-box hooks, with no access to weights or activations. It keeps a secret, distribution-fixed comparison core, whose frozen distribution makes its capability proxy comparable across generations, alongside a rotated fresh layer that hardens the bank against co-adaptation. Four tests built on that proxy cover the level gap, a scale-aligned divergence with online change-point detection, capability stagnation, and a conditional confidently-wrong rate; a Sidak correction turns them into a calibrated family-wise p-value. Diagnosis alone recovers nothing, so a risk-aware immunization layer reselects an honest candidate from the proposal pool using the core together with a purely structural gaming footprint, disclosing at most log2 Pi bits per generation to the host. We prove a detectability bound that converts a target error rate into an explicit probe-size budget, and we delimit what probe rotation does and does not buy. On a controlled prompt-level host with four injected hacking channels and ground-truth labels, HackProbe reaches 0.763 AUROC against 0.663 for the strongest baseline and cuts the false-positive rate from 0.706 to 0.434. Its bandwidth-limited reselection is the only immunization level that returns more true capability under hacking, 5.2 points on average, than it forfeits on clean runs, 4.7; per-channel effects are mostly not individually significant.
Read the full abstract

All authors Rongxin Yang, Yang Liu, Shang Luo, Haoxuan Jia, Chongyang Zhang, Hao Zheng, Yingguang Yang, Yulin Huang, Jianshen Zhang, Yongzhi Qi, Kefu Xu, Congjing Ran, Bin Chong

High Energy Physics (Theory)

Hironaka Geometry and Resurgence in Finite-$N$ Matrix Models

We study how finite-$N$ invariant theory organizes the non-perturbative structure of matrix models. For a model of four traceless Hermitian $2\times 2$ matrices, the Hironaka decomposition realizes the gauge-invariant configuration space as an eight-sheeted branched cover of the space of primary invariants. We show that ramification points of this cover naturally generate additional saddle points when the action depends only on the primaries. For an explicit rank-two saddle, we compute its action, one-loop normalization and Picard–Lefschetz connection to the perturbative vacuum. The same saddle controls the leading Borel singularity and large-order growth of perturbation theory, while its first fluctuation correction reproduces the first subleading large-order correction with no fitted parameters. Our results provide a concrete link between finite-$N$ invariant geometry and resurgence in matrix models.
Read the full abstract

Mathematical Software

BF16 Component-Product Emulation of FP32 and FP64 GEMM on Intel AMX

Modern CPUs increasingly integrate high-throughput matrix engines optimized for low-precision AI workloads, while many scientific computing applications still rely on FP32 and FP64 GEMM to meet their numerical accuracy requirements. This paper presents a method that combines BF16 matrix products on Intel Advanced Matrix Extensions (AMX) to emulate real square FP32 and FP64 GEMM. For FP32 GEMM, each element in both input matrices is decomposed into three BF16 components, and six selected component products are evaluated to target FP32-level accuracy comparable to oneMKL SGEMM. For FP64 inputs within the supported BF16 exponent range, the method uses a simplified fixed six-slice Ozaki decomposition. Each retained BF16 product is first produced in FP32, then widened and accumulated in FP64. Four product-count settings retain 6, 10, 15, or 21 component products, exposing the accuracy-performance tradeoff relative to oneMKL DGEMM. The FP32 schedule reuses component data and retains partial sums in AMX tiles. AMX-FP32 outperforms oneMKL SGEMM at tested orders $N\ge4096$ on 64 threads and also achieves gains for smaller matrices at lower thread counts. AMX-FP64 variants retaining fewer component products can outperform oneMKL DGEMM at sufficiently large orders; more products improve accuracy at additional cost.
Read the full abstract

Signal Processing

Is Forecasting Accuracy Enough? A Comparative Study of Traffic Forecasters for Beam-Hopping LEO Satellite Networks

We evaluate diverse models for user traffic demand forecasting in Low Earth Orbit (LEO) satellite networks with Beam Hopping (BH), questioning whether predictive accuracy is the right objective for this task. To capture the complex nature of the user traffic demand, we employ a second-order self-similar traffic model, supplemented by a publicly available Wi-Fi dataset to validate the self-similar model against the empirical traffic patterns. We compare forecasters ranging from classical statistical approaches, such as the optimal forecaster for self-similar data and the optimal linear predictor on the discrete sampling grid, to Fractional Auto-Regressive Integrated Moving Average (FARIMA) models, as well as emerging deep learning architectures. The latter category encompasses foundation and domain-specific transformer models, alongside a lightweight neural network consisting solely of linear layers. We assess these models at two levels: in isolation, through the Mean Absolute Scaled Error (MASE), and in context, through a BH simulator in which the forecast drives the illumination plan. On purely self-similar traffic the three self-similarity aware forecasters perform on par with one another and dominate the learned models, whereas on the raw Wi-Fi trace this ordering nearly reverses. Seasonality violates their stationary increment assumption; removing the periodic component restores their comparative accuracy. Crucially, these accuracy differences barely propagate to the system level. Loss ratio and buffer backlog are affected more by system utilization and the planning period than by the choice of forecaster, with the performance gap between forecasters vanishing entirely below 0.90 utilization. This suggests design efforts are better spent optimizing utilization margins and planning periods rather than chasing marginal gains in raw accuracy.
Read the full abstract

All authors Yekta Demirci, Guillaume Mantelet, Stéphane Martel, Jean-François Frigon, Gunes Karabulut Kurt

Machine Learning

Interpretability for Turing Machines

We show that susceptibilities, an interpretability technique developed for neural networks, can identify the presence of algorithmic structure in Turing machines by probing the local loss landscape of a learning problem for noisy Turing machines introduced by Murfet and Troiani (arXiv:2504.08075). We prove that symmetries and path separation in the algorithm implemented by a Turing machine induce permutation symmetries and low-rank blocks in its susceptibility matrix. We study this empirically on a set of deterministic finite automata (DFAs) and demonstrate that algorithmic features can be recovered by principal component analysis and clustering methods in susceptibility space.
Read the full abstract

All authors Billy Snikkers, Rumi Salazar, Daniel Murfet, Will Troiani

Algebraic Topology

Twisted Bicategorical Shadows and Traces

Bicategorical shadows provide a categorical framework that encompasses Hochschild homology and topological Hochschild homology (THH), encodes their Morita invariance, and extends the notion of trace from symmetric monoidal categories to bicategories. Certain equivariant variants, such as $C_n$-twisted THH, do not fit into the ordinary shadow framework. We introduce bicategories with $G$-twisting data and show that every ordinary shadow on such a bicategory induces, for each $g\in G$, a $g$-twisted shadow on the associated bicategory of $G$-twists. These $g$-twisted shadows are invariant under $G$-Morita equivalence; examples include $C_n$-twisted THH and twisted Hochschild homology of $C_n$-Green functors. We further define a $g$-twisted bicategorical trace that recovers the $g$-twisted Hattori–Stallings trace. For $C_n$-Green functors, the orbitwise $g$-twisted Hattori–Stallings traces assemble into a morphism of $C_n$-Mackey functors that, at the orbit $C_n/C_n$, recovers the degree-zero twisted Dennis trace.
Read the full abstract

Numerical Analysis

A Complete Resolution of Forsythe's Conjecture for Restarted Conjugate Gradients

Forsythe's conjecture, published in 1968, asserts that for each restart length $s$, every exact-arithmetic restarted conjugate-gradient iteration on a real symmetric positive definite problem either terminates or has normalised residuals that converge separately along the even and odd restart subsequences. Apart from the classical steepest-descent case, this asymptotic question remained unresolved in full generality for nearly six decades. We give a complete classification by restart length in the original finite-dimensional setting and identify a sharp threshold. For $s=2$ and $s=3$, every problem either terminates or has convergent even and odd residual directions. For every $s\ge4$, there is a diagonal positive definite counterexample of dimension $s+4$ which never terminates and whose even residual directions do not converge. Together with Akaike's theorem for $s=1$, this shows that the conjectured universal conclusion is true precisely for $s\in\{1,2,3\}$ and false for every $s\ge4$. The positive results follow from a degree-independent double-orthogonality identity and an analysis of the low-degree fixed-point sets. At restart length four, rational interval arithmetic and Sturm sequences certify a transverse Hopf point of the leading vector field of the rescaled squared-weight map. Analytic periodic-orbit and shadowing arguments yield the counterexample at restart length four, and degree elevation extends the construction to every larger restart length. The classification for all $s\ge2$ is also formally verified in Lean.
Read the full abstract

Information Retrieval

VizIt: A multi-view framework for exploring single-cell, spatial, and genetic data online

Multi-omic studies increasingly require data to be examined from complementary biological perspectives, yet interactive exploration remains fragmented across modalities and tools. We present VizIt, an open-source framework for multi-view exploration of single-cell and spatial transcriptomic, epigenomic and genetic data. VizIt connects gene-, cell type-, condition-, spatial-, genomic region- and variant-centered views, enabling seamless navigation across biological perspectives. We demonstrate VizIt through the Parkinson's Cell Atlas, a customizable interactive multi-omic resource.
Read the full abstract

All authors Chenhang Christopher Zhang, Yanqing Lou, Jie Yuan, Mingming Lu, Jacob Parker, Himanshu Chintalapudi, Zechuan Lin, Clemens R. Scherzer, Yuxuan Hu, Ruifeng Hu, Xianjun Dong

Combinatorics

The Weighted Walks in Quadrant with Finite Groups: an Algebro-Geometric Approach

We classify weighted small-step lattice walks in the quadrant whose associated birational group is finite. Using an algebro-geometric description of the kernel curves and Cayley-type finite-order conditions, we relate the group $G_W$ of a walk to the family of groups $\Gamma_t$ acting on its kernel curves. Together with the uniform upper bound on the order of $G_W$, this allows us to analyse all possible finite orders. We obtain explicit necessary and sufficient conditions for the group to have order 4, 6, 8, or 10, and prove that no weighted walk in the quadrant has a group of order 12. This yields a complete classification of weighted quadrant walks with finite groups.
Read the full abstract

Optics

High-Performance Photonic Crystal Biosensors using Sub-wavelength Nanopillars

Photonic-crystal (PC) slabs composed of subwavelength silicon pillars erected on a thin silicon backing layer, with heights of only one or half lattice constant, are simulated for optical biosensing by numerical solution of Maxwell's equations. Localized optical cavity modes that form waveguides in the direction of biofluid flow, with quality factors surpassing \(10^4\), are embedded within the 2D photonic band gap of the 3D structure. Remarkable sensitivity to analyte binding is achieved using fragmented thin silicon pillars in regions of maximum localized optical-field intensity. Exceptional quality factors are realized through destructive wave interference of vertical radiation loss. This provides a rare combination of high sensitivity and low limit of detection in a device with minimal time requirement for analyte binding. The sensitivity to analyte binding is a frequency shift of nearly \(0.004\,[2\pi c/a]\) per analyte-thickness increment \(0.1a\). The sensitivity to background biofluid change is about \(4800\) nm per refractive index unit (RIU) for a specific lattice constant \(a = 5\,\mu\mathrm{m}\) using light of wavelength \(16.5\,\mu\mathrm{m}\).
Read the full abstract

Materials Science

4 MV/cm (010) $\beta$-Ga$_2$O$_3$ Heterojunction Diodes Realized by Low-Damage e-Beam NiO$_x$ Interlayers

We report on the utilization of a low-damage e-beam NiO$_x$ deposition process to realize field-plated heterojunction diodes (FP-HJDs) on (010) $\beta$-Ga$_2$O$_3$ films with high critical breakdown field strengths beyond 4 MV/cm and power figure of merits (PFOM) of >1 GW/cm$^2$. Diodes were fabricated on a 2.67 $\times$ 10$^{16}$ cm$^{-3}$ intentionally doped 6.2 $\mu$m thick epitaxial layer grown by metalorganic chemical vapor deposition (MOCVD) on a conductive Sn-doped $\beta$-Ga$_2$O$_3$ (010) substrate using TMGa, O$_2$, Ar carrier gas, and SiH$_4$ as the silicon dopant source. The mesa etched FP-HJD devices utilized a thin 7 nm e-beam NiO$_x$ interlayer before the sputtered NiO$_x$ layers to eliminate the effects of sputter-induced ion damage on the (010) epilayers. Current-Voltage measurements resulted in a forward current density of 700 A/cm$^2$ at 4 V, HJD ideality factor of 1.29, a V$_{bi}$ of 2 V, rectification ratio of 10$^{11}$, and a differential specific on resistance (R$_{on,sp}$) value of 2.36 m$\Omega$ $\times$ cm$^2$. Breakdown of the (010) FP-HJDs was 1.64 kV, leading to a parallel plane electric field at breakdown (E$_{||,max}$) of 4.02 MV/cm and a PFOM of 1.14 GW/cm$^2$, which is a state-of-the-art result for diodes on MOCVD-grown (010) drift layers.
Read the full abstract

All authors Carl Peterson, Yizheng Liu, Chinmoy Nath Saha, Rachel Kahler, Akhila Mattapalli, Sriram Krishnamoorthy

Combinatorics

Random independent sets and local sparsity

We analyze random constructions of independent sets in locally sparse graphs, specifically graphs with bounded maximum average degree in neighborhoods or with fractionally $r$-colorable neighborhoods. Specializing our methods to finding large independent sets and low-weight fractional colorings, we focus on optimizing for marginals, but we also derive results that find many independent sets (i.e.\ give lower bounds on the independence polynomial) by optimizing for entropy. Our main results generalize the local Shearer bound of Martinsson and Steiner for triangle-free graphs to graphs with few triangles and to graphs with fractionally $r$-colorable neighborhoods, in the latter case improving upon a result of Dhawan. We also extend independence polynomial bounds obtained via induction to such graphs, improving upon known bounds obtained by local occupancy by relaxing the necessary hypotheses from a maximum degree condition to an average degree condition.
Read the full abstract

Computation and Language

Choosing the Right Language Mode at Inference Time for Multilingual Reliability

Multilingual large language models often struggle to reason in low- to mid-resource languages. Prior work has shown that translation can improve multilingual reasoning by helping models access stronger English-centric representations. This raises a central question: How much translation is needed for multilingual large language models to reason reliably, and when does more translation instead trigger interference and overconfidence? Using LLaMA and Qwen models, we run extensive experiments varying text scope and language mode (target-only, English-only, bilingual) to evaluate both accuracy and reliability. Our results reveal a clear trade-off: English context often improve understanding and recover errors caused by non-English comprehension, yet adding redundant bilingual context intensifies interference. We address this trade-off with Reliability-Aware Adaptive Inference (RAAI), a training-free test-time framework that (i) performs Expected Calibration Error (ECE)-aware routing and prompt fusion, and (ii) uses a mid-layer Risk Index (RI) to gate sequential reasoning, allocating compute only when it is likely to help and suppressing harmful bilingual redundancy. Across two model families, RAAI enhances accuracy by 25-37.7% on low-resource languages and lowers calibration error by approximately 3-6%, with the most pronounced benefits in the lowest-resource language tiers.
Read the full abstract

Quantum Physics

Qmes: Quantum Meta-Learning for Encoding Selection in Quantum Kernel Methods

Selecting an effective encoding quantum circuit is a key challenge in quantum kernel methods because different feature maps can lead to different performance. Conventional methods require constructing and evaluating every circuit for each new dataset, making it computationally expensive. We present Qmes, an open-source Python package that automatically recommends circuits through meta-learning. Qmes characterizes a dataset using classical complexity measures and queries a pre-trained model to recommend circuits without quantum evaluation at inference time. The package provides modular components for meta-feature extraction, quantum-kernel evaluation, recommender training, model selection, and user-defined circuit extension. We validate Qmes on 105 classification and 86 regression benchmark datasets. Qmes reduces the mean recommendation regret by 2.2x and 4.2x for classification and regression, respectively, compared to a non-adaptive baseline, with statistical significance confirmed via a paired Wilcoxon signed-rank test ($p < 10^{-4}$). Qmes thus enables efficient and practical encoding-circuit selection for quantum kernel methods.
Read the full abstract

All authors Dao Duy Tung, Quoc Chuong Nguyen, Vu Tuan Hai, Le Bin Ho, Lan Nguyen Tran

Artificial Intelligence

Continual Graph Memory for Adaptive Recommendation under Intent Drift

This paper studies adaptive recommendation under intent drift, where feedback from each recommendation outcome can reveal whether the relational evidence used for ranking is useful, missing, or misleading. While Knowledge Graphs (KGs) provide essential semantic structure to handle these shifts, traditional KG-enhanced systems treat the graph as a static retrieval substrate, making it brittle to evolving intents, noisy metadata, and recurring failure patterns. This paper proposes CGM-Rec, a continual graph memory framework for adaptive recommendation. CGM-Rec treats the graph state as a writable memory and maintains two complementary components. Therein, a Semantic Graph Memory is updated conservatively through quality-gated typed operations for storing stable and high-confidence relational knowledge. Meanwhile, an Episodic Lesson Memory acts as a fast reactive memory that learns recent outcomes, failure cases, and corrective hints. During testing, model parameters remain frozen and adaptation occurs only through memory writes. We evaluate CGM-Rec under a frozen-parameter, one-pass reranking protocol, where encoders and prompts remain fixed during testing and adaptation occurs only through memory writes. Experiments across multiple recommendation settings show that CGM-Rec improves over evaluated neural and LLM-based baselines on most metrics. Particularly, under sampled-candidate reranking, CGM-Rec improves HR@1 by up to 29.58% over the strongest LLM baseline on Bundle, and outperforms K-RagRec on metadata-rich ML-100K with HR@5 of 0.5941 versus 0.4746.
Read the full abstract

All authors Hao Nguyen Ngoc, Tung Nguyen, Nguyen Thi Hanh, Hoang Thai Dinh, Nguyen Xuan Tung

High Energy Physics (Phenomenology)

Self-Interacting Sterile Neutrino Cold Dark Matter: Resonant Production Mechanism in the Early Universe

Sterile neutrinos are well-motivated dark matter candidates, but their conventional production through active-sterile mixing is tightly constrained by X-ray searches and structure-formation observations. We propose a distinct production mechanism operating entirely within a sterile sector: two sterile neutrinos, $N_1$ and $N_2$, coupled to a singlet scalar $\phi$, with $N_1$ the dark matter candidate and $N_2$ held in equilibrium through frequent scattering induced by its scalar interaction. Thermal self-energies induced by the $N_2$ and $\phi$ backgrounds generate both a temperature-dependent mass splitting and an off-diagonal mixing between $N_1$ and $N_2$. As the Universe cools, the in-medium levels can undergo a level crossing, leading to resonantly enhanced conversion of the thermal $N_2$ population into $N_1$. We formulate the conversion using a density-matrix kinetic equation that consistently incorporates coherent $N_1$-$N_2$ conversion, collisional decoherence, and thermal repopulation of $N_2$. For a narrow resonance, the integrated conversion probability admits a simple analytic form that coincides with the Landau-Zener result, despite the underlying collisionally damped dynamics. In the weak-conversion regime relevant for freeze-in, this correspondence provides a robust analytic description of the resonant production. We derive the resulting dark matter abundance and identify the conditions for cosmological stability of $N_1$ and for resonant conversion to dominate over direct scattering and decay production. The resulting relic abundance scales as $Y_1\propto g_{12}^2 g_{22}M_{\rm Pl}/m_1$, making the dark matter energy density approximately independent of $m_1$. This mechanism provides a new route to sterile-neutrino dark matter that does not require appreciable active-sterile mixing.
Read the full abstract

Computer Vision

ReaDiT Guidance: Control for Image and Video Generation using Diffusion Transformer Features

We present DiT Readout (ReaDiT) Guidance, a lightweight framework for controlling generation with Diffusion Transformer (DiT) models via their internal feature representations. ReaDiT Guidance uses features from a single DiT block to steer the generative process according to spatial targets - like depth, pose, or edge maps - provided at test time. Furthermore, since modern text-to-video models are largely built on DiT backbones, ReaDiT Guidance naturally extends to video generation, enabling camera and motion control. Experimental results demonstrate that our approach achieves competitive or improved results compared to existing feature-based and off-the-shelf adapter-based approaches while requiring fewer parameters.
Read the full abstract

All authors Jay Mahajan, Chang Liu, Rauf Makharov, Viraj Shah, Alexander Schwing, Svetlana Lazebnik

Computation and Language

ConsensusBench: Benchmark of Consensus Nodes for LLM Reasoning via Outcome Reward Densifying

Reinforcement learning (RL) has become one of the primary paradigms for reasoning enhancement of large language models (LLMs). In particular, Group Relative Policy Optimization (GRPO) and related algorithms have demonstrated strong performance with outcome-level rewards. However, these methods depend solely on the final answer, without feedback regarding which intermediate steps contribute to success or failure. As task complexity and reasoning trajectory length increase, such sparse final-answer rewards become increasingly insufficient. To address this limitation, we introduce ConsensusBench, a novel dataset designed to provide rule-based process-level signals. We posit that a correct final answer relies on a small set of intermediate conclusions throughout the reasoning process, which can be seen as a verifiable sub-outcome. We identify these sub-outcomes by filtering correct trajectories from N rollouts and clustering semantically equivalent intermediate statements. We call these clustered statements as Consensus Nodes. By integrating a rule-based process reward derived from these nodes into GRPO-style algorithms, we develop a new reinforcement learning signal named ConsensusPR. It directly reduces the reward sparsity of outcome reward across long reasoning trajectories. To facilitate systematic process-level evaluation, we introduce three metrics to our benchmark: Final Answer Accuracy (Acc), Node Coverage Rate (NCR), and Tokens per Node (TPN). Experiments across AIME 2024, AIME 2025, GSM8K, MATH-500, and our ConsensusBench demonstrate that the proposed method consistently surpasses GRPO-style approaches, highlighting the practical value of consensus nodes in guiding reasoning.
Read the full abstract

All authors Shi-Qi Yan, Chao-Hong Tan, Qian Chen, Wen Wang, Xiangang Li, Zhen-Hua Ling

Computation and Language

CAGE: Coherence-Aware Graph Encoding for Retrieval-Augmented Generation

Traditional Retrieval-Augmented Generation (RAG) systems score each passage independently against the query, assembling context sets that may be individually relevant yet collectively incoherent. We introduce Coherence-Aware Graph Encoding (CAGE), a reranking framework that models "between-chunk coherence" across four dimensions: Intra-Domain Relevance, Noise Resistance, Informational Bonding, and Factual Consistency. Our pipeline transforms retrieved passages into directed heterogeneous entity graphs, amplifies factual anchors via min-out-degree reweighting, encodes structural patterns through a Relational Graph Convolutional Network, and fuses inter-chunk coherence with query relevance for final ranking. Evaluated across four multi-hop benchmarks, CAGE matches or outperforms strong baselines including monoT5 in Recall@5 on bridge-dominated datasets and consistently improves downstream Exact Match, demonstrating that structurally coherent context yields more precise answers even when retrieval recall is comparable or lower.
Read the full abstract

All authors Tong Qi, Jingyu Wu, Youbing Yin, Spencer Hong, Daben Liu, Erin Babinsky

Computer Vision

Importance-Aware Low-Rank Distillation of Diffusion Transformers

Diffusion Transformers (DiTs) have emerged as a dominant architecture for high-quality text-to-image generation, yet their scale poses challenges for efficient deployment. While truncated singular value decomposition (SVD) is a principled tool for parameter reduction, evidence from large language models (LLMs) suggests that naive low-rank approximation can cause catastrophic failure. In contrast, we find that truncated SVD in DiTs produces smooth degradation even under substantial global compression, with redundancy distributed across projection matrices throughout the whole network rather than concentrated in a few transformer blocks. Building on these insights, we introduce SVDtrunc, a two-step block-level compression scheme, first allocating ranks across blocks and compressing the least important ones via truncated SVD under a global parameter budget, and then fine-tuning all blocks with modular knowledge distillation and a rectified-flow objective. We apply SVDtrunc to FLUX.dev across compression levels ranging from 40-90% of the original parameter count. Across three benchmarks, GenEval, HPSv2, and DPG, we outperform all competing approaches. Notably, and in contrast to prior work, we retain near-full performance at 68% and remain competitive even at 57% of the original parameter budget. Furthermore, we show that SVDtrunc complements step distillation and achieves strong results even without fine-tuning, positioning it as a practical continuation of efficiency improvements beyond diffusion step reduction for large-scale generative models. Project page: https://vislearn.github.io/SVDtrunc/
Read the full abstract

All authors Denis Zavadski, Sebastian Heid, Damjan Kalšan, Stefan Roth, Carsten Rother

Information Retrieval

Latent-Aligned Reasoning for Multimodal Recommendation

Multimodal Vision-Language Models (VLMs) have demonstrated remarkable capabilities in cross-modal understanding, yet a fundamental challenge persists when applying them to recommendation: as representations propagate through multi-step reasoning, both visual and textual signals progressively attenuate - a phenomenon we term cross-modal dilution. To address this, we propose LARK (Latent-Aligned Reasoning frameworK), a two-stage latent reasoning framework with complementary alignment mechanisms within a single VLM. In the first stage, learnable latent tokens are interleaved with multi-step chain-of-thought (CoT) reasoning and explicitly aligned with a frozen vision encoder, serving as visual checkpoints that preserve perceptual details throughout the reasoning chain. In the second stage, the latent representations are projected via a bridge MLP and trained with item-to-item contrastive learning; to prevent the reasoning semantics from fading, intermediate features are aligned with the CoT hidden states from the first stage, anchoring the final embeddings to the model's own reasoning output. Experiments on three public benchmarks and one industrial dataset show that LARK achieves state-of-the-art performance across multiple recommendation architectures, with controlled ablations confirming the distinct contribution of each component.
Read the full abstract

Logic

Constructive equivalence between Brouwer's fixed-point theorem and weak König's lemma

In the context of constructive reverse mathematics, we show that Brouwer's fixed-point theorem and weak König's lemma (WKL) are equivalent. To derive WKL from Brouwer's fixed-point theorem, the construction of a continuous function on the unit square without fixed points due to Orevkov [Soviet Math. Doklady (1963), 1253–1256] is generalised to yield a uniformly continuous function on the unit square whose fixed points encode information about infinite paths of a given infinite tree.
Read the full abstract