Computer Vision
Papers
Fresh from arXiv, typeset properly. Highlighted words trade on the market.
All of arXiv
Tuesday 1 September
Eye-tracking data are expensive to collect, requiring specialized hardware and controlled laboratory conditions, and difficult to share due to privacy regulations, which together limit the quantity and diversity of gaze data available for research. We investigate unconditional gaze generation using diffusion models in position and component-wise velocity representations. The models are trained and evaluated on eight-second visual-search trajectories from 28 participants, using a split of 23 training participants and five held-out validation participants, with three training seeds. Evaluation includes kinematic distributions, trajectory geometry, event statistics, and a temporal Wasserstein GAN with gradient penalty as a learned comparator. Across eight position-derived features, mean Jensen-Shannon divergence is $0.016 \pm 0.004$ for the position model and $0.037 \pm 0.002$ for the velocity model, with standard deviations calculated across training seeds. These scores do not establish that representation choice caused the difference. A separate comparison using the same evaluation method shows that the ordering of the models varies across training seeds. The position model also produces smaller feature discrepancies than a kinematic Markov baseline. Differences in saccade counts between real and generated data depend strongly on the velocity threshold used to detect events. The results provide quantitative reference measurements for unconditional gaze generation on this visual-search task and show how evaluation choices affect conclusions about generated gaze.
Robotics
Situation Awareness for Intelligent Data Distribution in Connected Vehicles
The limitations of on-board sensors and blind spots caused by occlusion cause the reduction of perception quality in autonomous vehicles. In such cases, cooperative perception provides additional data via Vehicle-to-Everything communication to enhance local perception, causing a large volume of data transmission. The vehicle can focus on acquiring and utilizing relevant data according to the prevailing road context by identifying the current traffic situation. To achieve this, we propose a concept for the situation identification of the vehicle using Bird's-Eye-View images. Firstly, the situation around the vehicle is identified using object detection with semantic segmentation, followed by understanding the context of the traffic using a situation identification module consisting of an open-source projective transformation network Cam2BEV and a situation identification neural network. The concept was evaluated and validated by running the software on the CARLA simulator using the in-built RGB camera and the semantic segmentation camera. Additionally, the portability of the situation identification module for real-world applications was verified on Cityscapes and nuScenes urban driving datasets. Overall, the proposed situation identification approach enables efficient sensor data management by prioritizing relevant data to the current traffic situation. The source code is available in the following link: https://github.com/akshaynarla/DySi_Select
Read the full abstract
Computer Vision
Contrastive Knowledge Distillation for Anomaly Detection in Multi-Illumination/Focus Display Images
In this paper, we tackle automatic anomaly detection in multi-illumination and multi-focus display images. The minute defects on the display surface are hard to spot out in RGB images and by a model trained with only normal data. To address this, we propose a novel contrastive learning scheme for knowledge distillation-based anomaly detection. In our framework, Multiresolution Knowledge Distillation (MKD) is adopted as a baseline, which operates by measuring feature similarities between the teacher and student networks. Based on MKD, we propose a novel contrastive learning method, namely Multiresolution Contrastive Distillation (MCD), which does not require positive/negative pairs with an anchor but operates by pulling/pushing the distance between the teacher and student features. Furthermore, we propose the blending module that transforms and aggregate multi-channel information to the three-channel input layer of MCD. Our proposed method significantly outperforms competitive state-of-the-art methods in both AUROC and accuracy metrics on the collected Multi-illumination and Multi-focus display image dataset for Anomaly Detection (MMdAD).
Read the full abstract
Geometric Topology
Minimal-Volume Equiangular Hyperbolic $4$-Polytopes
We study finite-volume convex hyperbolic $4$-polytopes whose dihedral angles are all equal to a fixed strictly acute angle $\alpha$. Put $\alpha_0=\arccos(1/3)$ and $\alpha_1=\arccos(1/4)$. We first show that the class is empty for $\alpha<\alpha_0$. For $\alpha_0\leq\alpha<\alpha_1$, the unique polytope of minimum volume is the regular hyperbolic $4$-simplex with dihedral angle $\alpha$. As $\alpha$ increases to $\alpha_1$, this simplex degenerates to a Euclidean one and no hyperbolic simplex exists at $\alpha_1$. For $\alpha_1\leq\alpha<\pi/2$, the unique minimum is the regular equiangular hyperbolic $4$-cube. The proof combines the hyperbolic Gram–Euler relation with the Davis–Okun theorem on the Charney–Davis inequality for flag triangulations of the $3$-sphere. The geometric step is a missing-face argument showing that the boundary of the dual of every nonsimplex in the class is flag.
Read the full abstract
Combinatorics
A Counterexample to Teschner's Bondage-Number Conjecture
For a finite simple graph $G$ with at least one edge, the bondage number $b(G)$ is the least number of edges whose deletion increases the domination number $\gamma(G)$. Teschner conjectured that $b(G)\le \tfrac32\Delta(G)$ for every graph $G$. We disprove this conjecture by giving a connected cubic bipartite graph on eighteen vertices with \[ \gamma(G)=6 \qquad\text{and}\qquad b(G)=5. \] The domination number is established by a complete counting argument across the bipartition. An explicit five-edge deletion raises the domination number from six to seven. For the matching lower bound, we give an exact finite certificate: the graph has 297 minimum dominating sets, and deleting any one of its $\binom{27}{4}=17{,}550$ four-edge subsets leaves at least one of those sets dominating. The enumeration is deterministic, uses only exact integer and set operations, and is reproduced by the complete standard-library verifier included in the appendix.
Read the full abstract
Audio and Speech Processing
Probing Warmth-Mediated Harm in Speech-Enabled LLMs for Mental-Health Conversations
Audio LLM benchmarks measure understanding and dialogue quality, not whether speech-enabled models respond with relational warmth when a vulnerable user discloses a mental-health concern. We introduce a 7-turn scripted-disclosure probe grounded in WHO mental-health clinical guidelines, with each script run on the same model (Azure OpenAI gpt-realtime) in both audio and text-only conditions, and acoustic-prosody analysis of the generated speech. Across 532 responses we identify two audio-specific patterns transcript-only evaluation would miss: at the elicitation turn the model's voice gets shorter, faster, lower-pitched, and quieter rather than warmer (p < .001 for five of seven acoustic features), and the modality gap on relational acceptance, small in aggregate, concentrates in the highest-stakes self-harm/suicide scripts. A two-rater listener study corroborates that perceived warmth is concentrated at specific turns and on bereavement disclosures. Together these patterns indicate that auditing speech-enabled models in mental-health contexts requires evaluating the combined audio-and-text experience the user encounters, not the transcript in isolation. We release the protocol, scoring pipeline, and scripts as a starting point for evaluating speech-enabled models in mental-health contexts.
Read the full abstract
Information Retrieval
SAGE: Semantic Attribute Graphs for Multi-Entity Visual Retrieval
Dense document images often contain many fine-grained visual and textual entities whose relevance depends on a user query. Standard vision-language retrievers encode cropped regions with a single vector, which can mix distinct entity signals and obscure the evidence needed for fine-grained retrieval. We call this failure mode Semantic Dilution and quantitatively show that it degrades entity-level retrieval as a function of entity density. To mitigate it, we propose SAGE, a training-free framework that parses semantic entities from dense document images, represents them as hierarchical graph nodes with multi-vector embeddings, and retrieves query-relevant evidence through iterative entity-level subgraph matching. We also introduce DEAR, a dataset of 1,055 query–image pairs sourced from product detail pages, where each query requires retrieving and comparing multiple fine-grained entities from visually dense inputs across four question types of increasing complexity. Experiments show that SAGE substantially reduces semantic dilution and outperforms patch-level and OCR-based retrieval baselines on DEAR, achieving a Recall@3 of 0.849 and a generation score of 2.746 on multi-entity visual comparison queries. Our code is available at https://github.com/All4Nothing/SAGE.
Read the full abstract
General Relativity and Quantum Cosmology
Fixed ADM reconstruction of extended uncertainty principle black holes
We study a phenomenological fixed-ADM prescription for an extended-uncertainty-principle (EUP) correction to the Schwarzschild exterior. Rather than treat mass constancy as a consequence of the EUP, we hold the charge measured at spatial infinity fixed so that we can compare exteriors at the same physical mass. We impose the EUP-shifted horizon radius and Hawking temperature and select the slowest admissible inverse-power representative. Its Einstein tensor defines an effective anisotropic tensor, but we do not infer an independent matter model, quantum state, or perturbation theory from that definition. For a positive EUP parameter, the effective energy density is negative throughout the exterior, while the transverse null and strong energy conditions fail over an extended radial region. The area entropy and the formal entropy obtained from the reduced relation $\mathrm{d}M=T\,\mathrm{d}S$ disagree at leading order; we interpret this mismatch as evidence that the thermodynamic description lacks work terms or an action-based entropy law, not as a definite EUP entropy correction. An explicit one-parameter family with the same mass, horizon, and temperature shows that the orbit, lensing, and test-field eikonal shifts depend on the chosen representative. Our results therefore characterize the minimal member of a phenomenological family rather than universal consequences of the EUP.
Read the full abstract
High Energy Physics (Experiment)
Comment on arXiv:2608.29450: "First Measurement of Solar Neutrinos through Elastic Neutrino-Electron Scattering at the keV Scale"
The XENONnT Collaboration recently reported a new measurement of the solar \(pp\) neutrino flux via neutrino-electron elastic scattering using a multi-ton-scale liquid xenon detector. We comment here on an important omission in their summary plot.
Read the full abstract
History and Philosophy of Physics
The Whittaker-Maxwell Scaffolding. Scaffolding or template: what remains of a model when it is removed
This paper discusses the metaphor, originating with Whittaker, according to which Maxwell's model of molecular vortices was the scaffolding by means of which the equations of electromagnetism were erected, removed afterwards without consequence for the building. It is argued here that the metaphor describes badly what happened, and it is shown that whoever upholds it today transcribes, within that same text, the quotations that belie it. The paper examines what was actually removed in 1865, which materials of the model still operate inside the equations under other names, and proposes in its place the figure of the template, taken from sheet-metal layout work. The discussion is situated with respect to the work of Daniel M. Siegel and Cameron Lazaroff-Puck, who maintain from the history of science the constitutive character of the model. It closes with a criterion for telling one case from the other.
Read the full abstract
Materials Science
FrOGS: Discrete Neural Sampler for Independent Alloy Configurations Across Chemical Conditions
Predicting the thermodynamic properties of an alloy requires sampling its configurations across many chemical conditions and recovering free energies on a common absolute scale. Markov chain Monte Carlo (MCMC) is the standard tool, but it requires separate simulations at different conditions, and auxiliary free-energy methods such as thermodynamic integration are used to place results on a common absolute scale. Modern discrete neural samplers typically use reverse KL divergence as the objective and can be mode-seeking or biased. We present Free energy Offering Generative Sampler (FrOGS), a hybrid discrete neural sampler that couples an autoregressive model to a continuous-time Markov chain (CTMC) to be trained jointly under a single shared loss. FrOGS draws i.i.d. configurations, returns an unbiased estimate of the partition function, and gives consistent estimates of thermodynamic observables. We train a single model across a wide range of chemical conditions to produce estimates on a common absolute free-energy scale. FrOGS matches exact finite-size results on the 2D Ising model and reference phase diagrams for AgPd and CuAu, without mode collapse. We additionally compare to SEGAL, a published autoregressive baseline, and find that only FrOGS recovers the stability range of the CuAu$_3$ phase.
Read the full abstract
Cryptography and Security
Privacy-Preserving Heterogeneous Multi-LLM Federated Inference for Cognitive Diagnosis
Significant challenges remain in AI-driven educational systems in balancing privacy preservation with accurate cognitive diagnosis. To overcome this, we propose a federated inference framework in which several commercial LLM APIs collaborate without requiring access to raw student data or proprietary model internals. Using multiple federated entities, such as LLaMA-3.3-70B, GPT-4o-mini, and Claude-3-Haiku, our framework builds upon a heterogeneous multi-LLM architecture. The predictions generated by these entities are combined with epsilon-local differential privacy by adding Laplace noise locally to each entity's prediction output before aggregation, while residual-based aggregation mitigates model heterogeneity. Our approach is predicated on an honest-but-curious trust paradigm in which API providers are presumed not to abuse submitted queries, and our differential privacy mechanism shields the published diagnostic results from external inference. We conduct rigorous privacy-utility analysis showing strong privacy guarantees with minimal accuracy loss, and extensive real-world evaluations across three educational benchmarks confirm the framework's practical usability and cross-domain generalizability.
Read the full abstract
Mesoscale and Nanoscale Physics
A Unified Description of Dirac-Cone Anisotropies in Two Dimensions
Anisotropies in two-dimensional materials are responsible for a variety of effects that significantly modify their physical properties. In this paper, we present a covariant modification of the Dirac equation that incorporates anisotropies into the effective low-energy description around the Dirac point. The model is constructed by analogy with a Lorentz-violating extension of the Standard Model of elementary particles and yields a (2+1)-dimensional framework describing physical effects such as shifted, tilted, and distorted Dirac cones through its free parameters. The proposed model thus provides a general and unified framework in which distinct anisotropy-induced modifications of the Dirac spectrum and their combinations can be systematically characterized. We further apply the model to strained graphene and show that its effective parameters can be quantitatively extracted from first-principles electronic band structures while retaining a direct geometrical interpretation. These results establish a connection between the microscopic electronic structure and a material-independent effective description of anisotropic two-dimensional Dirac systems.
Read the full abstract
Group Theory
On embeddings of the difference graph of the intersection power graph and the power graph
The power graph of a finite group $G$ is a simple undirected graph with vertex set $G$ and two vertices are adjacent if one is a power of the other. The intersection power graph of a finite group $G$ is a simple undirected graph with vertex set $G$ and two vertices $x$, $y$ are adjacent if $\langle x\rangle \cap \langle y \rangle \neq \{e\}$. The difference graph $\mathcal{D}(G)$ of a finite group $G$ is the difference of the intersection power graph and power graph with all isolated vertices removed. We characterized all the finite nilpotent groups $G$ except $2$-group such that the difference graph is planar. Further, we determine all the finite nilpotent groups whose difference graph has genus at most $2$. Moreover, we prove that there does not exist any finite group whose difference graph is projective planar.
Read the full abstract
Information Retrieval
Reflect-SQL: A Self-Reflection Based Framework for Text-to-SQL
Democratizing data access through natural language is a crucial goal for modern enterprises, but the practical adoption of Text-to-SQL is critically hindered by real-world complexities: 1. Obscure and large database schemas, 2. Ineffective retrieval of relevant tables and columns due to structured setting of schemas and vague user query, 3. Generation of syntactically or logically flawed SQL due to a lack of robust validation and correction mechanism. To address these systemic challenges, we introduce Reflect-SQL, a novel framework for Text to SQL, grounded in multi-stage self-reflection approach to develop understanding of obscure schema using a knowledge base, setup a process for effective retrieval and system to generate syntactically/semantically SQL. Instead of a single-pass attempt, our system employs an LLM-as-a-judge driven scoring mechanism within interconnected feedback loops to iteratively refine the results at every stage. A feedback-driven retrieval loop refines the user's natural language query, while a synthesis loop validates and corrects the SQL and finally, an entailment loop optimizes the end-to-end process and continuously enriches the knowledge base. By integrating these layers of reflection, Reflect-SQL bridges the critical gap between user intent and complex data. On the challenging BIRD benchmark, our framework achieves an execution accuracy of 72.03%, significantly outperforming state-of-the-art baselines, demonstrating a major leap in reliability for enterprise applications.
Read the full abstract
Fluid Dynamics
Super-Resolution Initialization of High-Fidelity CFD Simulations for Pebble-Bed Reactors
High-order CFD simulations provide detailed resolution of the heterogeneous interstitial flow in pebble-bed reactors, but their computational cost is high, especially during the initial flow-development period required to reach statistically stationary conditions. This work investigates the use of a Super-Resolution Graph Neural Network (SR-GNN) to improve the initialization of high-order NekRS simulations. Lower-order P = 2 velocity fields are used as inputs to reconstruct higher-order representations, which are then used as initial conditions for P = 7 restart simulations. The approach is evaluated using a 146-pebble bed at Re = 1000, Re = 2500, and Re = 5000, with pressure-drop convergence used as the main figure of merit. The SR-GNN models were trained using paired low- and high-order snapshots and were first evaluated through qualitative inference comparisons. High-order restart simulations showed that, for Re = 1000 and Re = 2500, the SR-GNN initialized cases produced pressure-drop histories similar to direct restarts from true P = 2 fields. For Re = 5000, however, the super-resolved field restart approached the statistically stationary P = 7 pressure-drop range faster than both the direct P = 2 restart and the reference P = 7 simulation initialized from a uniform velocity field. The trained Re = 5000 model was also applied to a larger 1568-pebble bed, demonstrating qualitative applicability of the workflow to a significantly larger packed-bed geometry. These results indicate that SR-GNN-based initialization is a promising strategy for reducing high-order flow-development cost, while also motivating further work on broader Reynolds-number and geometry generalization.
Read the full abstract
Computer Vision
Physics-Driven Independent Pair Generation for Iterative Self-Supervised Low-Dose CT Denoising
Low-dose computed tomography (LDCT) measurements contain mixed Poisson-Gaussian noise. However, most self-supervised methods rely on generic image statistics and do not explicitly model this noise, which may limit their ability to effectively suppress realistic LDCT noise. To address this issue, we propose a physics-driven framework with cross-domain iteration for self-supervised LDCT denoising. The proposed framework proceeds in three main steps. First, a learned sinogram prior and the LDCT noise model guide posterior inference of photon counts, enabling separation of the Poisson and Gaussian components. Second, the separated Poisson and Gaussian components are respectively processed by binomial thinning and Gaussian data thinning to construct two branches, and residual scaling matches each branch's noise level to that of the observation, yielding a training pair with approximately independent noise realizations from one low-dose measurement. Finally, the pair is used to train an image-domain network whose forward-projected outputs update the prior. Through cross-domain iteration, the prior and the training pair are progressively refined while maintaining consistency with CT acquisition physics. Experiments on simulated data from AAPM, LIDC-IDRI, and LoDoPaB-CT and on real LDCT data show consistent gains over the evaluated self-supervised baselines across dose levels, with performance comparable to the evaluated supervised baseline.
Read the full abstract
Mathematical Physics
Superintegrability of stratified symplectic spaces
We define the notion of superintegrability of a Hamiltonian system on a stratified symplectic space. We focus on spin Calogero-Moser-Sutherland (sCMS) systems, where the phase space is a stratified symplectic space obtained by the Hamiltonian reduction of the cotangent bundle over a compact Lie group, and demonstrate that the sCMS systems for $SU(3)$ are superintegrable.
Read the full abstract
Machine Learning
Convergence Theory of Knowledge Distillation in Asynchronous P2P Gossip Learning Network
Decentralized, serverless learning increasingly connects devices running different architectures, where the standard tool, decentralized SGD, is undefined as models with different parameter counts cannot be averaged. Knowledge distillation (KD) exchanges soft predictions rather than weights and sidesteps this obstacle, yet convergence theory for fully decentralized, asynchronous peer-to-peer (P2P) KD is lacking. We provide one, relocating consensus from parameter space to function (output) space: a KD event is a geometric contraction operator in logit space on the peers' predictive distributions, which we analyse in the Hilbert space of predictions on a reference measure. Under standard smoothness/variance assumptions and two realizability assumptions, one bridging parameter SGD to the functional step and one controlling restricted task/KD alignment, the time-averaged functional stationarity and function-space disagreement converge at rate $O(1/(\eta T))$ to an $O(\eta)+O(B_f^2)+O(\zeta_f^2)$ neighbourhood. Here $B_f$ is the distance from the task optimum to the peers' reachable classes and $\zeta_f$ measures persistent local-task heterogeneity. Across homogeneous, width-heterogeneous, and mixed-family networks of the experiments, KD contracts function disagreement by $40-61\times$, while isolated training does not. The sampled stationarity diagnostic has late transient exponents $0.99-1.90$ on the shared-skeleton main runs, and the four-point step-size sweep exhibits the predicted transient: neighbourhood tradeoff.
Read the full abstract
Quantum Physics
Exact diffusion coefficients for quantum–classical hybrid walks
One-dimensional quantum–classical hybrid walks are introduced by combining quantum and classical steps under periodic and random protocols. Exact diffusion coefficients are derived for both protocols throughout the diffusive regime. The exact results show how diffusion depends on the average frequency and temporal arrangement of classical steps. As the average frequency approaches zero, both diffusion coefficients diverge inversely with the frequency. At matched average frequencies, the ratio of the diffusion coefficients for the random and periodic protocols approaches two. A finite-mode approximation is also developed and tested against the exact diffusion coefficient for the random protocol. The exact results for the one-dimensional model provide a foundation for future studies of more general quantum–classical hybrid walks.
Read the full abstract
Statistical Methodology
Bigraphical Matérn-Whittle (BMW) Processes for Fast Inference of Big Multivariate Spatial Data on General Domains
Large spatial data sets now record many correlated variables at many thousands of locations, often on domains where Euclidean distance misrepresents proximity. The central difficulty is modelling the cross-variable dependence jointly while retaining variable-level interpretation. We introduce the bigraphical Mat'ern-Whittle process, a multivariate Gaussian process that resolves this with two graphs. A spatial graph generates the Mat'ern structure of each variable through a fractional power of a graph Laplacian, so the process is valid on any topology, with per-variable range, smoothness and amplitude. A directed acyclic variable graph encodes the scientific structure: we prove that each absent edge yields an exact conditional independence between the corresponding fields. We further prove that the operator determinant does not involve the cross-dependence coefficients, which keeps matrix-free likelihood evaluation and Bayesian learning of the variable graph tractable at scale. Estimation requires only sparse matrix-vector products and scales to tens of millions of space-variable pairs. In simulations the method recovered parameters and graphs accurately, remained robust under misspecification, and halved held-out prediction error on a non-convex domain. In a spatial transcriptomics section with 19,809 cells and 1,122 genes, fitted in 75 minutes on a laptop, borrowing across the learned gene graph reduced held-out prediction error by 50 to 91 percent. Theoretical challenges, such as the achievable efficiency of estimating the variance of the nugget, are also explored.
Read the full abstract
Data Structures and Algorithms
Forbidden Subgraphs of Graphs with Low Bandwidth
A layout of a graph G is an injective function $f : V(G) \rightarrow Z$, and the bandwidth of a layout f is $bw(G,f) = max_{uv \in E(G)} |f(u) - f(v)|$. The bandwidth bw(G) of G is the minimum bandwidth of a layout of G. Computing the bandwidth of a graph is a notoriously hard problem: assuming P != NP, there is no polynomial time algorithm, even on very restricted classes of trees [Monien, SIAM Journal on Algebraic Discrete Methods, 1986], and no constant factor approximation, even on trees [Dubey et al., JCSS 2011]. Assuming the Exponential Time Hypothesis, there is no algorithm with running time $f(k)n^{o(k)}$ to determine whether an input graph has bandwidth at most k, even on very restricted classes of trees [Dregi and Lokshtanov, ICALP 2014]. In this paper we show that Bandwidth on general graphs is FPT-approximable. In particular we give an algorithm that takes as input a graph G and an integer k, runs in time $2^{O(9^k)}n^{O(1)}$, and outputs a subtree T of G such that $bw(T) \geq k$ or a layout of G of bandwidth at most $(10^{85} k^{28})^{4^k}$. This resolves in the affirmative an open problem of Chung and Seymour [Discrete Mathematics, 1989], who asked whether the bandwidth of every graph G is upper bounded in terms of the maximum bandwidth of a subtree of G. Our theorem leads to a forbidden subgraph characterization for graphs of bounded bandwidth, and can be seen as an analog for bandwidth of the classic grid minor theorem for treewidth, the forbidden subtree theorem for pathwidth, and the forbidden subpath theorem for treedepth.
Read the full abstract
Hardware Architecture
An Emerging NVM-Based On-Chip Training Architecture with Non-Ideality Mitigation Through Bipolar Weight Distributions
The rapid advancement of deep learning has presented significant energy efficiency challenges to the conventional von Neumann architecture. In-memory computing (IMC) architectures based on emerging non-volatile memory (eNVM) are widely regarded as a promising solution for accelerating neural network training due to their high parallelism and low power consumption. However, the intrinsic non-idealities of eNVM devices can cause conductance updates to deviate from target values, thereby limiting the performance of on-chip training. To address this challenge, this paper presents a Non-ideality Optimized eNVM Accelerator (NOVA) architecture for on-chip training. Specifically, we first fabricate a two-dimensional (2D) ferroelectric field-effect transistor (FeFET) and develop a conductance modulation behavioral model calibrated with experimental data. Building upon this device model, we propose, for the first time, a Non-ideality Avoidance Training (NAT) algorithm tailored for eNVM devices, which mitigates accuracy degradation by guiding weight convergence toward the most stable conductance regions of eNVM devices. Experimental results demonstrate that, even under severe device asymmetry, NAT improves the accuracy by an average of 15.1% over the baseline methods across multiple benchmark tasks. Meanwhile, the NOVA achieves an average energy efficiency gain of approximately 33.58$\times$ compared with the peak energy efficiency of graphics processing units (GPUs).
Read the full abstract
Machine Learning
On-Policy Distillation Meets Off-Policy GRPO: Training Compact Instruction-Following Rerankers
Compact instruction-following rerankers are attractive for deployment, but conventional distillation pipelines typically train students by offline imitation of teacher outputs on a fixed set of examples, constraining supervision to the teacher's observed ranking space. We revisit reranker distillation through the lens of reinforcement learning. We propose a two-stage framework combining off-policy teacher optimization with on-policy student distillation. In Stage 1, a 4B teacher reranker is strengthened with off-policy GRPO using LLM-judge feedback on 88K instruction-following examples. In Stage 2, a compact 1B student samples rankings from its own policy and receives soft teacher-derived rewards on those rankings, coupling student exploration with knowledge transfer. Our strongest gains appear under distribution shift. On MAIR-11, the original 11-subset, 869-query evaluation, the proposed student reaches 0.7670 nDCG@6, outperforming offline listwise KD by +4.6 points. Controlled comparisons against offline pairwise RankNet KD and on-policy GKD show that neither changing the offline distillation objective nor moving teacher-distribution matching on-policy reproduces the performance of reward-based on-policy distillation over student-sampled rankings. The advantage persists on MAIR-Full: across all 126 tasks and 9,356 queries, the proposed method obtains the highest task-macro point estimates among the evaluated distillation variants, reaching 0.6808 nDCG@6 and 0.7865 MRR@6. It also exceeds two released 7B RL-trained rerankers on the comparable MAIR-11 evaluation, while the same Stage 2 training procedure consistently improves three architecturally distinct alternative student backbones. On the 9,861-query validation benchmark, the resulting 1B reranker achieves 0.7624 nDCG@6 while providing a favorable quality-efficiency tradeoff relative to larger alternatives.
Read the full abstract
Biological Physics
Antipolar Cell-cell Adhesion-causing Collective Motility Disorder
In this study, we aim to theoretically investigate antipolar cell-cell adhesion, in which adhesion sites are located on the opposite side of the leading edge of migrating cells, as a candidate for irregularly polarized adhesion that induces disorder in collective cell migration. We employ the cellular Potts model to simulate the effects of antipolar adhesion on collective migration driven by cell motility. Antipolar adhesion induces a collective motility disorder, which exhibits a disordered configuration in the motility direction, even when collective motion occurs in the absence of adhesion. Consequently, antipolar adhesion inhibits collective migration. The effect is in contrast to that of polar adhesion, which accelerates the directional intercellular order of cell motility. At a specific motility strength, a depinning transition emerges from a collective motility disorder to a collective motion. The collective motility disorder can be physically explained by the cooperative effect between antipolar adhesion and motility persistence within the mean-field approximation.
Read the full abstract
Cryptography and Security
Pushing Forward Multi-Secret-Key Homomorphic Encryption for Private Average Aggregation
Federated Learning enables multiple clients to train a shared model while keeping their local datasets isolated. However, the exchanged model updates may still leak sensitive information, making private aggregation a central building block in practical deployments, especially in the cross-silo setting. Homomorphic Encryption naturally fits the client–aggregator communication pattern of Federated Learning, but conventional single-key deployments rely on strong non-collusion assumptions. Multiparty Homomorphic Encryption removes this limitation, although recent attacks under restricted decryption access require large-variance smudging noise during collaborative decryption, which significantly increases ciphertext size and implementation complexity. In this work, we propose lightweight multi-secret-key protocols for private average aggregation based on RLWE-based Homomorphic Encryption. Our construction departs from the usual multiparty blueprint by avoiding the generation of a collective public key. Instead, each client encrypts its update under its own secret key, while the resulting ciphertexts remain compatible with homomorphic aggregation and collaborative decryption. By explicitly tracking and cancelling the ciphertext noise during decryption, the protocol removes the need for large $\lambda$-dependent smudging noise. We instantiate the construction with both exact BFV-based and approximate CKKS-based variants, prove its security in the semi-honest model against an adversary corrupting the aggregator and up to $L-1$ clients, and compare its communication and runtime performance with state-of-the-art MHE-based aggregation. Our results show that the proposed approach substantially reduces ciphertext expansion and online cost, while preserving practical homomorphic aggregation performance.
Read the full abstract
Cryptography and Security
Privacy Amplification Without Independence: How Far Negative Dependence Carries the Guarantees of Poisson Subsampling
Poisson subsampling is the default sampler in differentially private optimization because its independence makes privacy amplification tractable. Practical systems, however, are moving toward structured participation: random allocation (balls-in-bins), per-epoch allocation, random check-ins, schemes widely believed to be at least as private as Poisson subsampling at the matched rate. We isolate the probabilistic mechanism behind this belief and delimit it exactly, for Gaussian mechanisms up to correlated-noise matrix mechanisms. (1) If the participation indicator vector is negatively associated (NA), then at every integer Rényi order $\alpha\ge2$, exactly at all finite parameters, its remove-direction Rényi divergence is dominated by that of the marginal-matched independent scheme. For fixed gradient sequences, this extends to the mechanism level whenever the noise strategy's Gram matrix is sign-balanced, an $O(t^2)$-checkable condition. (2) The integer-order restriction is essential. For random allocation with $k=1$, we prove a linear law for the Rényi-difference criterion: at large $t$, dominance reverses for every $\alpha<3/2$, including KL divergence, while the crossing order tends to $3/2$ independently of $\sigma$. (3) We also localize the known failure of rate-matched Poisson domination exactly: below $(1-q)^t$, the hockey-stick ordering reverses, so substituting the Poisson pair into composition machinery is unsound. An upper-tail argument yields a finite crossover $\gamma_\star$, connecting this threshold picture to the Rényi boundary at $3/2$. Together, these results give a substitution map for privacy accounting: when Poisson-based computations remain sound for structured participation, where they fail, and what sound alternatives cost in deployment.
Read the full abstract
Econometrics
Headline Estimation with Multiple Research Designs
To study a scalar parameter, a researcher may consider multiple research designs. Based on the evidence across designs, the researcher may wish to formulate a headline estimate of the parameter. I examine how to choose this headline when it is unclear which design is most appropriate for studying the parameter. I model this setting by assuming that (i) exactly one of the designs is valid for the parameter and (ii) the researcher has ambiguity about which design is valid, represented by a class of priors over the candidate designs. To account for ambiguity, I propose reporting the headline estimate that minimizes the worst-case posterior risk over the class of priors. In three applications, I show cases where accounting for ambiguity materially affects the quantitative conclusion and cases where an existing headline is already close to optimal.
Read the full abstract
Machine Learning
Refining Heuristic-Based Bitcoin Address Clustering with Graph Neural Networks
Bitcoin's pseudonymous nature makes it challenging to analyze user-level activity, since a single user may control multiple identifiers (addresses). Existing heuristic-based methods attempt to identify addresses belonging to the same user, but they often produce flat cluster assignments with limited modularity and are prone to errors such as merging different users together. In this work, we propose a method for refining heuristic-obtained clusters by grounding our clustering on contrastive embeddings yielded by graph neural networks. Our contributions are threefold: (i) we release a publicly available dataset of Bitcoin transaction graphs containing a substantial number of clusters; (ii) we propose a methodology for learning address embeddings consistent with heuristics, and back it up with theoretical guiding intuitions; (iii) through hierarchical clustering, we enable a finer analysis of heuristic clusters and provide a quantitative criterion for flagging suspicious merges.
Read the full abstract
Statistics Applications
Neural Networks Learning the Radon–Nikodym Derivative: Empirical Option Pricing in Incomplete Markets
In incomplete markets, no-arbitrage (NFLVR) guarantees the existence, not the uniqueness, of an equivalent local martingale measure (ELMM): unhedgeable risks (jumps, stochastic volatility) admit a whole family of equivalent measures, and asset dynamics alone cannot pin down the one the market selects. We characterize the identifiability of $Q$ from option data and propose a measure that is identifiable from data yet prices any claim consistently. The key boundary is an “identification wall”: European options identify only the terminal marginal, while out-of-sample tails and path/joint structure require instruments matched to the priced risk (variance or higher-moment swaps, path-dependent claims). Within this view, minimum-relative-entropy weighted Monte Carlo (WMC) is the optimal baseline for the marginal; we generalize it to a full path-space measure change parameterized by a neural network on physical scenarios—XiNet—which learns $\xi=\mathrm{d}\mathbb{Q}/\mathrm{d}\mathbb{P}$ directly from 8 model-free path features, with option prices as a soft constraint. On European (marginal) pricing XiNet matches but does not surpass WMC, and both are bound by the identification wall out-of-sample. On path-dependent claims the picture reverses: calibrated on the same European surface, per-marginal methods fail structurally (an ATM forward-start is mispriced by $\sim+100\%$), whereas XiNet's single self-consistent measure keeps the bias to $+0.3\%$, beating maximum-entropy WMC ($-24\%$), because $\xi=f_\theta(\text{path features})$ captures joint structure Europeans cannot constrain. Identification is thus risk-specific, and XiNet is a single measure that absorbs available instruments and prices all claims consistently.