Peer Revenue Returns subject to peer review

Papers

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

All of arXiv

Tuesday 10 November

Robotics

DeRep: Learning Depth Representations for End-to-End Navigation in Cluttered Environments

End-to-end navigation in cluttered environments relies on compact visual representations to support collision avoidance and goal-directed motion. Existing methods typically learn these representations jointly with the policy or through observation reconstruction, which does not explicitly specify what information the representation should preserve. Moreover, they are usually evaluated with a single downstream policy learner, although the benefit of a representation may depend on how the policy is learned. It therefore remains unclear how different representation-learning objectives affect downstream navigation performance and robustness. This work proposes DeRep, a navigation-oriented representation-learning method that combines local bird's-eye-view reachability supervision with depth reconstruction. To compare representation objectives under controlled conditions, we pretrain encoders with a common architecture and freeze them during downstream learning. We evaluate the representations using a linear policy probe, flow-matching imitation learning, and reinforcement learning with PPO. DeRep achieves the lowest linear policy probe prediction error and outperforms reconstruction alone across all six tested PPO conditions. In imitation learning, both reachability-based objectives achieve higher mean success rates than reconstruction alone, with similar mean success rates between the reachability-only and joint objectives. These results demonstrate consistent benefits of incorporating reachability supervision into depth representation pretraining across different downstream learning methods and evaluation conditions. We further deploy the PPO policy with the frozen DeRep encoder in Gazebo and on a physical quadrotor without fine-tuning, demonstrating its capability for zero-shot sim-to-sim and sim-to-real transfer.
Read the full abstract

Machine Learning

Evaluating Compositional Generalization in Transformers: The Role of Composition Equivalence and Module Coverage

Compositional generalization is an important form of out-of-distribution generalization, defined as the ability to train on some combinations of modules (e.g., functions such as sort, filter) and then generalize to unseen combinations. Despite substantial empirical work evaluating compositional generalization in transformers, findings have been inconsistent because evaluations were often conducted on a limited set of train-test splits, defined in an ad hoc manner. Through a systematic evaluation, we identify two key aspects of the train-test distribution that strongly affect performance: composition equivalence and module coverage. We show that the apparent performance of direct models (trained only on final outputs) results primarily from exploiting composition equivalences, meaning different sequences of modules that reduce to identical end-to-end functions. When benchmarks eliminate these equivalences, performance drops significantly, showing their inability to generalize to compositions of known modules that produce novel end-to-end functions. We further demonstrate that performance degrades monotonically as module coverage shifts increase (when modules appear in different contexts between train and test), affecting both direct and step-by-step models (trained on intermediate and final outputs). These findings have concrete implications for benchmark design: benchmarks that include compositional equivalences can lead to significant overestimation of the generalization abilities of direct models, and benchmarks that include both module coverage violations and compositional equivalences can exacerbate shortcut learning. As a case study, we demonstrate the role of compositional equivalence on the generalization performance of the pre-trained model Llama3-8B on the code-output prediction task in the coding benchmark LiveCodeBench.
Read the full abstract

All authors Purva Pruthi, Andrew Yuan, Alexander D'Amour, David Jensen

Systems and Control

Impact of Reduced-Order Modelling on Sum-of-Squares Region-of-Attraction Certification for Grid-Forming and Grid-Following Converters

With the increasing penetration of power-electronic equipment, the transient stability of grid-forming (GFM) and grid-following (GFL) converters is increasingly characterized through the region of attraction (ROA) of the post-fault equilibrium. Sum-of-squares (SOS) programming yields certified inner approximations of the ROA, yet its computational cost grows rapidly with system order. Reduced-order models are therefore widely used to reduce computational cost and, more importantly, to enable analytical investigation. However, ROA analyses based on these models often assume that reproducing the converter's transient response, or retaining more dynamics in the correct time-frame, is sufficient to preserve the ROA of the full-order system. Using time-domain ODE simulations as a reference, the SOS-based ROA of each reduced-order GFM and GFL model is assessed against both its own simulated ROA and that of the corresponding full-order model, thereby separating SOS-induced conservativeness from model-order-induced ROA distortion. The results indicate that increasing reduced-order model complexity does not necessarily improve SOS-based ROA fidelity. In the cases examined, more complex reduced-order models yield lower SOS coverage of the full-order reference ROA, despite closer ODE-boundary agreement in some regions. Model selection should therefore consider both full-order ROA fidelity and certified coverage of the practical operating range, rather than model complexity alone.
Read the full abstract

Computer Vision

Reasoning-Aware Relational Representation Learning for Open-Vocabulary Scene Graph Generation

Open-Vocabulary Scene Graph Generation (OV-SGG) requires models to recognize visual relationships beyond the training vocabulary, yet existing methods often rely on dataset-specific object–predicate co-occurrence patterns. Under long-tailed distributions and noisy object localization, such reliance leads to biased, entangled relation representations, limiting both rare-relation recognition and generalization to unseen predicates. To address these challenges, we propose RaRe, a reasoning-aware relational representation learning framework that transforms MLLM-generated chain-of-thought (CoT) rationales into structured relation embeddings. Rather than using CoT only as an intermediate reasoning trace for final prediction, RaRe directly optimizes the rationale-derived representation space to improve inter-relation discriminability. Specifically, RaRe first aligns rationale embeddings with textualized relation descriptions through self-supervised contrastive learning, and then refines rationale generation with a specificity-oriented reinforcement learning objective that encourages semantically object-pair-specific relational evidence. Experiments on VG150 and PSG demonstrate improved category-balanced recognition over existing baselines, while the additional gains from learning vary across datasets and proposal settings. Code available at: https://artanic30.github.io/project_pages/RaRe.
Read the full abstract

All authors Jiawei Xiao, Boya Wang, Longtian Qiu, Shan Ning, Xuming He

Combinatorics

The Largest Ground Set for Sunflower-Free Systems

We construct the largest $r$-flower-free system of $k$-uniform sets for $k=3, r=4$ where every element of the ground set $G$ is used at least once. The goal is to maximize $|G|$, also called the span. We define $\mathrm{sp}(k, r)$ to be the maximum $|G|$ for such a system. We show that $\mathrm{sp}(3, 4) = 39$, which yields an improved upper bound $f(3, 4) \leq 130$, improving upon the Erdős–Rado bound of $f(3, 4) \leq 162$.
Read the full abstract

Robotics

Route-Time Conflict Maps for Cooperative End-to-End Autonomous Driving

Existing cooperative end-to-end driving works at Class A (status-sharing) of the SAE J3216 classes of cooperative driving automation (CDA): vehicles exchange what they perceive, not what they intend, so each plans alone when to go, which limits cooperation and leads to deadlock when both vehicles yield or to collisions when both go. We present When2Go, a cooperative end-to-end driving system that learns when to go by sharing intent under CDA Class B or higher. Each vehicle broadcasts its LiDAR detections with forecast motion as instance tokens, together with its plan and intended trajectory. The receiver draws the fused detections, forecasts and partner plans into a route-time conflict map, the occupancy along its route over the planning horizon, on which a planner scores candidate speed profiles. Where two intended trajectories conflict, an order gate, a precedence network distilled from a right-of-way rule, decides who goes first; both vehicles evaluate it on the same messages and so reach the same order. When2Go imitates the rule-based expert of V2Xverse, which sees the simulator's ground truth; stops that When2Go cannot observe within its 2 s planning horizon are relabelled, so that it does not learn to wait without an observable cause. For our runs only, the benchmarks' PID controller steers by When2Go's waypoints instead of the route, and its forced push of long-waiting vehicles is off. With one model, When2Go sets the state of the art on three closed-loop CARLA benchmarks: 97.8 and 90.7 driving score on InterDrive without and with traffic (best published 90.1 and 78.9), 70.0 on the public V2Xverse protocol (best of sixteen rerun methods 59.1) and 88.2 on MDrive (best rerun 85.1), with 95.0% of its vehicles collision-free (87.7%).
Read the full abstract

All authors Haohua Que, Mingkai Liu, Tianle Zhu, Zhipeng Bao, Yiping Li, Handong Yao

Analysis of PDEs

On the two-component $b$-family system modeling shallow-water waves of medium amplitude

We formally derive a two-component $b$-family system from the full governing equations for two-dimensional water waves in the medium-amplitude regime. The system contains higher-order nonlinearities and includes a two-component Camassa-Holm type system for equatorial shallow-water waves with the Coriolis effect as a special case. We establish the local well-posedness of the Cauchy problem in $B^{s-1}_{p,r}\times B^s_{p,r}$ with $s>\max\{1+\frac1p,\frac32\}$. Using Littlewood-Paley theory and Osgood's lemma, we also prove local well-posedness in the critical Besov space $B^{\frac12}_{2,1}\times B^{\frac32}_{2,1}$. Furthermore, we derive a blow-up criterion for strong solutions. For a special choice of the parameters, we prove finite-time blow-up for a class of initial data and provide an upper bound for the maximal existence time.
Read the full abstract

Statistical Methodology

Pseudo-marginal Sampling for Selective Inference

Inference after adaptive selection must account for selection bias. One approach is to condition on the selected event, but for complex selection procedures, the resulting conditional distribution can be analytically intractable and require sampling-based methods. Rejection sampling provides a natural baseline: it repeatedly draws from a proposal and retains only samples that reproduce the observed selection. This can be prohibitively expensive when the selection event is rare. We address this computational bottleneck with a pseudo-marginal Markov chain Monte Carlo algorithm that samples from the selective distribution. Using a weak Poincaré inequality, we analyze its computational efficiency relative to rejection sampling and identify when it can offer substantial gains. In the classical file-drawer problem, we derive precise computational costs for both methods, illustrating exponential improvement as the selection event becomes increasingly rare. In numerical experiments involving complex selection procedures without tractable conditional distributions, the pseudo-marginal sampler is often hundreds of times more efficient than rejection sampling, while the resulting tests achieve nominal selective Type I error.
Read the full abstract

Cryptography and Security

Data-Efficient Unlearning-Based Backdoor Detection in Federated Learning

Detecting backdoor attacks in federated learning (FL) is challenging because malicious clients can manipulate the global model while preserving main-task performance. Static model-update statistics may therefore fail to clearly distinguish malicious from benign clients. Unlearning-based detectors instead probe client models through loss ascent on clean server data, but detection performance degrades with scarce proxy data. We propose Screening via Unlearning Response Analysis (SURA), a data-efficient detector. SURA reuses the same small clean proxy across multiple train-mode loss-ascent steps, strengthening client responses without additional clean data. For federated vision models with batch normalization (BN), repeated probing also changes BN running statistics and can confound endpoint measurements. SURA addresses this by restoring the shared pre-probe BN state only at readout while preserving the parameter changes induced by probing. This separates repeated probing from response measurement and removes BN drift and the interaction between parameter and BN-state changes from the readout. Across six backdoor attacks, SURA uses only 10 clean server images, outperforms methods under the same data budget, and approaches the robustness obtained with much larger proxy sets.
Read the full abstract

All authors Dowon Kim, Junhao Cai, Sung Il Choi, Dohun Kim, Juhyun Park, Chengjun Jin, Changhee Joo

Signal Processing

Design and Over-the-Air Evaluation of a Digital Twin-Driven Site-Specific Neural Receiver for 5G NR on NVIDIA DGX Spark

We design and evaluate a site-specific neural receiver for the 5G New Radio (NR) uplink. A corridor digital twin built from a radar scan and refined manually provides 28,671 ray-traced channel records for training an eight-round convolutional graph neural network with pilot-aware inputs and auxiliary channel supervision. The GPU on NVIDIA DGX Spark evaluates time–frequency convolutions in parallel across the received grid to accelerate inference. The receiver uses the standard NR uplink waveform and reference signals and feeds soft bits to the existing transport-block decoder. In over-the-air tests with NI USRP B210 radios under common radio settings, the site-trained receiver achieves the highest mean uplink rate at all five positions. At a non-line-of-sight corner, it reaches 16.2 Mb/s, compared with 4.1 Mb/s for the conventional receiver and 2.4 Mb/s for the 3GPP-trained neural receiver. These results demonstrate the feasibility of integrating neural reception into the 5G NR physical layer and show its potential for AI-RAN deployments.
Read the full abstract

All authors Qiuyuan Chen, Chengyang Yao, Liuchang Zhuo, Yuquan Sun, Cunhua Pan, Hong Ren, Jiangzhou Wang

Machine Learning (Statistics)

Simultaneous Near-Optimal Region Identification in Continuous Simulation Optimization: Minimax Complexity and Adaptive Sampling

Simulation optimization typically returns a single estimated optimal point. When decisions also depend on criteria evaluated independently of simulation, such as resource cost, the decision maker instead needs the set of designs whose expected performance lies within a tolerance of the optimum, often choosing the tolerance only after seeing the tradeoffs. We study how to identify these near-optimal regions on a continuous design space, simultaneously for all tolerances up to a prescribed maximum $T$, with region estimation error at most $\eta$. We show that region identification is not a by-product of optimization. Specifically, when $\alpha>\nu$, any procedure, even one given the optimal design in advance, needs at least $\Omega\bigl(\eta^{-(2+d/\nu)}\min\{1,KT^{d/\alpha}\}\bigr)$ simulation evaluations, where $d$ is the dimension, $\nu$ the smoothness, and $(K,\alpha)$ the covering parameters that describe the geometry of the near-optimal designs. We propose SASORI, which returns an estimated optimal point and the regions for all tolerances from the same evaluations, matching this lower bound up to logarithmic factors. SASORI also achieves the optimal tradeoff between region estimation error and cumulative regret, and reduces to a minimax-optimal simulation optimization procedure at zero tolerance. Numerical experiments illustrate the effectiveness of our proposed method.
Read the full abstract

All authors Yilin Zheng, Haowei Wang, Wenjia Wang, Szu Hui Ng

Computer Vision

When Reconstruction Helps Event-Camera Person Triggers

Event cameras record asynchronous changes in brightness. To use a general visual model as a recording trigger, a system must represent these measurements in a form the model can interpret. We compare direct grayscale accumulation, two temporal encodings, and two intensity reconstructors on paired real event data. On 192 windows from four held-out DSEC recordings, thresholds calibrated only on development data yielded 40 of 52 person detections and 12 of 140 background positives with E2VID. The strongest direct temporal encoding at an admissible fixed cutoff yielded 14 detections and seven background positives. A retrospective comparison at a common 10% background-positive cap yielded 40 detections for E2VID and at most 19 for any direct encoding. FireNet+ had lower preparation cost and produced 21 detections with zero observed background positives at its frozen cutoff. On the same 192 windows, Qwen3-VL-4B yielded 33/52, 2/140 with E2VID and 9/52, 9/140 with the highest-recall direct rendering, the three-bin image (person detections, background positives). A 96-second exploratory replay illustrates the sensitivity of episode capture and recording starts to cutoff selection. The experiments identify a detection benefit from reconstruction under controlled input history and distinguish it from the cost and false starts of continuous triggering.
Read the full abstract

Artificial Intelligence

The WoW: A Large-Scale World Knowledge Corpus for Full-Lifecycle LLM Training

Existing Chinese educational corpora often suffer from fragmented coverage and a lack of systematic organization. In this paper, we propose WoW (World Knowledge), a large-scale dataset designed for the full-cycle training of foundation models. The construction of this dataset is driven by an Educational Knowledge Graph, which serves as a taxonomic backbone for the systematic collection and completion of missing educational modules across all levels. Compared to previous versions, this iteration provides enhanced systematicity, superior quality, and a larger scale, while incorporating multimodal signals and supporting every stage of model development. Experiments demonstrate that WoW significantly improves model performance across diverse benchmarks. To facilitate community research, we further provide an interactive platform for graph-based data selection and retrieval.
Read the full abstract

Computer Vision

Hand-Object Contact Transition Detection in Egocentric Video

This paper tackles the task of hand-object contact transition detection in egocentric video, which aims to localize the two temporal boundaries of a contact episode: the moment a hand first establishes contact with an object (touch) and the moment contact is released (release). We propose a unified framework that jointly models the interacting hand and object to accurately localize both transition types. The proposed framework augments hand-centric spatiotemporal representations with object-aware features learned directly from the video and introduces a transition-aware event spotting formulation tailored to contact transitions, without relying on external object or hand-object interaction detectors during inference. To support this task, we present ContactTransition, a dataset comprising 4,021 egocentric videos with 8,498 touch and 8,077 release annotations, together with object-of-interest supervision. Experiments demonstrate that the proposed method consistently outperforms strong two-stage and end-to-end event-spotting baselines on both transition types. Finally, we show that detected contact transitions provide effective temporal anchors for downstream egocentric hand-object reasoning.
Read the full abstract

All authors Huy Anh Nguyen, Chang Dong, Feras Dayoub, Minh Hoai

Machine Learning (Statistics)

Differentially Private Statistical Inference for Rankings from Pairwise Comparisons

Reliable inference for rankings of large language models based on human pairwise comparisons must preserve the privacy of individual preference records. To meet this need, we develop methods for score and rank inference under the Bradley-Terry model with edge differential privacy. For an objective-perturbed estimator, a new leave-one-out analysis establishes finite-sample bounds on Euclidean and entrywise estimation errors. For sufficiently dense comparison graphs, the error bounds extend to smaller privacy budgets, including regimes in which its privacy term dominates its statistical term. A uniform asymptotic expansion yields asymptotically valid confidence intervals for individual scores and pairwise score differences. To improve finite-sample performance, we account for both regularization bias and privacy-induced randomness by combining a plug-in ridge-bias correction with Gaussian-Laplace calibration. We further establish a bootstrap procedure for simultaneous rank inference, providing asymptotically valid confidence sets for individual ranks and tests of top-$K$ membership. When the comparison probability is public and known, all proposed inferential procedures require no additional privacy budget. We evaluate finite-sample performance through simulations and apply the proposed methods to rank large language models for creative writing.
Read the full abstract

High Energy Physics (Theory)

Boundary dualities of symplectic Chern-Simons theories and minimal models

We propose that a Neumann-type boundary condition of the 3d $\mathcal{N}=2$ $USp(2n)$ Chern-Simons theory at the level $-(n+\frac32)$ with a fundamental chiral multiplet is dual to a Dirichlet-type boundary condition of an Abelian rank-$0$ theory. In a topological twist we propose that it supports the affine vertex operator algebra (VOA) $L_n(\mathfrak{osp}(1|2))$. With the corresponding specialization of the fugacities, the line defect half-indices of the antisymmetric Wilson lines give, up to monomials, the supercharacters of all its modules, which coincide up to signs with the characters of the Virasoro minimal model $M(2,2n+3)$. We extend the duality to the level $-(n+\ell+\frac12)$ and propose that the boundary VOA is $L_n(\mathfrak{osp}(1|2\ell))$. Together with the Giveon-Kutasov type duality and the mirror symmetry, the proposed dualities form a web of dualities of the boundary conditions and the line operators. For other gauge groups with real matter we propose that the boundary VOA in the holomorphic twist is the commutant of the gauge currents in free real fermions, which is the Ising model for $Spin(7)$ with the spinor and the $\mathcal N=1$ superconformal minimal model $SM(3,5)$ for $G_2$ with $\mathbf7$.
Read the full abstract

Chemical Physics

Coupled Cluster Simulations of Vibrational Circular Dichroism Spectra

We report the first simulations of vibrational circular dichroism (VCD) spectra at the coupled-cluster level of theory. We generalize Stephens's magnetic-field perturbation theory framework for the magnetic-dipole vibrational transition moment to biorthonormal wave functions. The resulting atomic axial tensors (AATs) and velocity-gauge atomic polar tensors collapse to Stephens's expressions for real, Hermitian states. Adapting these exact-state expressions to coupled-cluster left- and right-hand biorthonormal wave functions and using our recently developed density-based formulation yields compact equations that involve only perturbed ground-state amplitudes, as well as unperturbed and perturbed, unrelaxed one-electron densities; two-electron-density contributions vanish exactly. We have validated the resulting AATs by comparison with our previous configuration interaction formulation for a two-electron system, size-intensivity tests, a translational sum rule, and other tests. Simulated VCD spectra for (S)-3-chloro-1-butyne at the coupled-cluster singles and doubles level with the correlation-consistent triple-zeta basis set reveal significant discrepancies between length and velocity gauges. We find that the addition of large-exponent functions via a core-valence basis set for the Hartree–Fock reference components yields substantially improved gauge invariance without the need for gauge-including atomic orbitals.
Read the full abstract

Robotics

ExRL: Sample-Efficient Online Adaptation of Long-Chunk Policies via Execution Length Control

Predicting long action chunks, often spanning dozens of timesteps, has become a key ingredient in increasingly capable robot policies. In this paper, we find that adapting these long-chunk policies with existing reinforcement learning (RL) methods can require many online samples due to the high dimensionality of the action chunks being optimized, making such adaptation impractical in real-world scenarios where interaction data are costly to collect. Our key insight is to shift adaptation from optimizing actions to optimizing how long to execute each predicted action chunk. This keeps the adaptation space low-dimensional while its expressivity grows with the chunk size. We thus introduce Execution at the Right Length (ExRL), a simple approach which augments the base policy with a lightweight adapter that predicts the execution length of each chunk and is trained using off-policy RL algorithms. Across simulation benchmarks and real-world robotic tasks, ExRL enables sample-efficient adaptation of a diverse range of long-chunk policies with minimal task-specific hyperparameter tuning.
Read the full abstract

All authors Xu Luo, Jiaying Yang, Hao Wu, Junlin Xie, Zijin Hong, Ji Zhang, Lianli Gao, Jingkuan Song

Information Theory

Euclidean Hulls of Double Cyclic Codes of Arbitrary Block Lengths

Gao et al. \cite{GWF23} recently explored the Euclidean hulls of double cyclic codes over $\mathbb{Z}_2$. While they found answers for non-separable codes with coprime lengths ($\gcd(r, s) = 1$), the non-coprime case ($\gcd(r, s) > 1$) remained an open question. This paper completely solves that problem for all possible block lengths. We calculate the exact generator polynomials for the hull and show the precise conditions needed for a non-separable code to turn into a separable one. Using this, we show how to construct non-separable LCD codes. We also include several numerical examples to confirm our findings.
Read the full abstract

All authors Awadhesh Kumar Shukla, Om Prakash Pandey, Sachin Pathak, Ashish Upadhyay

Machine Learning

Sharp Sample Complexity of KL-Regularized Reinforcement Learning

KL-regularized reinforcement learning, the objective behind maximum-entropy RL and RL from human feedback, is variational inference in the graphical model of "control as inference": the regularized optimal policy is a posterior over trajectories, and the suboptimality of a policy is $\beta$ times a KL divergence between path measures. However, it is not known how many samples it takes to learn this posterior, even with a generative model. Earlier $1/\epsilon$ guarantees scale with $H^5$ in the effective horizon $H$, and no lower bound for MDPs depends on the temperature $\beta$. In this work, we show that for a discounted MDP with $K$ state-action pairs, the plug-in estimator finds a policy that is $\epsilon$-optimal from every initial state with $\widetilde{O}(KH^3/(\epsilon\max\{\epsilon,\beta\}) + KH)$ samples, and that no algorithm, even one that chooses where to sample, can do with fewer when $\epsilon \lesssim \min\{H, H^2/\beta\}$. For $\epsilon < \beta$ this is a $1/\epsilon$ rate with the same horizon dependence as unregularized RL. We also find the exact asymptotic constant of the plug-in estimator, a Fisher-weighted variance of the estimated soft Q-function. The estimator is locally asymptotically minimax among methods that sample every pair equally often, and we give the allocation of samples that is optimal among adaptive designs. Simulations on small MDPs match the rates and the constants.
Read the full abstract

Combinatorics

Log-Concavity in Narayana Transforms and Stack-Sorting Fibers

Given a permutation $\pi$ of size $n$, let $D_\pi(x)$ denote the descent polynomial of the fiber of $\pi$ under West's stack-sorting map. We prove that $D_\pi(x)$ is ultra-log-concave of order $n-1$ and, therefore, log-concave. More generally, we prove that a multivariate refinement of $D_\pi(x)$ recording ascent tops and descent tops is the basis generating polynomial of a laminar matroid. We deduce a result about simplex integration from Brändén and Huh's theory and apply it to weighted matchings, obtaining a refinement of the log-concavity of $D_\pi(x)$ that incorporates the peak statistic on the stack-sorting fiber of $\pi$. These results imply the log-concavity of certain sums of products of Narayana polynomials over the lattice points of integral binary nestohedra. A complementary argument shows that replacing the monomials of a homogeneous real-stable polynomial with nonnegative coefficients by products of shifted Narayana polynomials produces a real-rooted polynomial; we discuss several applications. The results in this paper were formalized in Lean by AxiomProver.
Read the full abstract

Spectral Theory

Half-inverse problems for perturbed Bessel operators with a known left-hand potential

For the regular Sturm-Liouville operator $-\frac{{\rm d}^2}{{\rm d}x^2}+q$ acting in $L^2(0,1)$, the Hochstadt-Lieberman theorem states that one full spectrum together with the potential on either the left half-interval $q|_{(0,1/2)}$ or the right half-interval $q|_{(1/2,1)}$ uniquely determines the full potential $q$. The inverse-square singular term $\tfrac{l(l+1)}{x^2}$ in the perturbed Bessel operator $-\frac{{\rm d}^2}{{\rm d}x^2}+\frac{l(l+1)}{x^2}+q$, acting in $L^2(0,1)$, breaks the left-right symmetry. It is known that, when the right half of the potential $q|_{(1/2,1)}$ is given, the uniqueness of the potential can be established even if a finite number of eigenvalues (depending on $l$) are omitted. In this paper we study the opposite setting–where the left half potential $q|_{(0,1/2)}$ is known–and show that one full spectrum together with finitely many norming constants uniquely determine the full potential $q$. Moreover, we specify the minimum number of norming constants required for the uniqueness. Finally, for positive integer $l$, $L^2$ potentials, and the Dirichlet boundary condition at $x = 1$, we establish the local bi-Lipschitz stability and the local solvability from the full spectrum and finitely many norming constants, assuming the potential on the left half-interval is fixed.
Read the full abstract

All authors Xin-Jian Xu, Chuan-Fu Yang, Natalia Bondarenko, Runpeng Xia

Machine Learning

Leveraging Psychophysical Attentional Distribution for Gaze-Augmented Reward Modeling

Integrating human feedback into language models has gained increasing attention as a means to align model outputs with human preferences, with Reinforcement Learning from Human Feedback (RLHF) serving as a prominent framework for preference-based alignment. Recent studies have incorporated eye-tracking (ET) data as an additional supervisory signal for reward model training, grounding reward learning in human reading behavior. Building on this approach, we extend prior fixation-centric models by leveraging established findings from cognitive psychology on human letter recognition. Specifically, we model visual attention as a spatially graded distribution centered on the current gaze location and incorporate this representation into reward model training. Our results show that the reward model trained with the proposed gaze distribution achieves higher preference prediction accuracy than a baseline model. Moreover, some results from our experiments are qualitatively consistent with established findings on human attention during reading, suggesting that they potentially capture cognitively meaningful aspects of human attention allocation.
Read the full abstract

All authors Chaeho Lee, Junyup Kim, Wansoo Kim, Young Min Jung, Sang Ho Lee

Optics

Hopfion Superlattices in 3D Twisted Moiré Light Fields

Moiré superlattices have recently been extended from electronic materials to photonic systems, enabling complex topological light fields such as skyrmion bags. However, these realizations remain confined to two dimensions. Here, we demonstrate 3D moiré superlattices formed by twisting multi-volume spatiotemporal optical lattices, giving rise to previously undiscovered 3D quasiparticle topologies, including hopfion bags, hopfion quasicrystals, and hopfion-hopfionium nested superlattices. By tailoring the interference polarization textures of the moiré spacetime lattices, we achieve controlled topological transitions among these states, and tailored topological numbers of hopfion bags. These spatiotemporal supertopologies offer a promising platform for high-capacity information encoding in higher dimensions.
Read the full abstract

Medical Physics

Quantum-walk transport oracles and amplitude estimation of a scalar range statistic in proton therapy

Quantum amplitude estimation (QAE) promises a quadratic reduction in the sampling cost of Monte Carlo (MC) particle transport, but two obstacles have kept it from radiotherapy physics: the lack of an efficient unitary encoding of charged-particle transport, and the mismatch between QAE's single-scalar output and the per-voxel structure of dose calculation. We address both. We construct a discrete-time quantum-walk transport oracle for protons in which an energy register is driven by an energy-dependent coin encoding a continuous-slowing-down surrogate of the stopping power, so that stochastic energy loss and range straggling follow from the circuit dynamics. The oracle is unitary, with one fresh coin qubit per step, and its gate count grows linearly in the number of energy levels. On it we formulate a scalar range statistic, the probability that a proton penetrates beyond a distal depth threshold, as a directly marked amplitude, and estimate it by maximum-likelihood amplitude estimation (MLQAE) with Grover iterations and finite shots. The oracle reproduces a Bragg-shaped depth-deposit profile agreeing with a classical MC of identical rules to within 0.27%, and Grover amplification matches $\sin^2((2k+1)\theta)$ to machine precision up to $k=32$. At matched budget the MLQAE error falls as $N^{-0.93}$ against $N^{-0.49}$ for sampling, 4.8 times smaller at the largest budget. A three-gate resource analysis (output structure, amortization, error-corrected wall-clock) finds full dose maps structurally unfavourable irrespective of hardware, and the scalar statistic short of a wall-clock crossover at clinical scale: the oracle's derived cost of $1.0 \times 10^6$ logical gates per call ($3.5 \times 10^7$ with rotation synthesis), dominated by the multiplexed coin, is 160 to 5400 times the permitted budget. We report the reduction that would be required and identify the term responsible for it.
Read the full abstract

Chaotic Dynamics

Unified Ott–Antonsen Framework for Stability and Bifurcations in Ring Swarmalators

Swarmalators, a prototypical model that integrates temporal rhythms with spatial swarming, have been widely used to explore the collective dynamics observed in active-matter systems. Although numerical simulations have revealed a variety of collective states, a general theoretical framework for the stability and bifurcations of the states remains unavailable. Here, by the model of one-dimensional ring-structured swarmalators with general distributions of natural frequencies and velocities, we conduct a systematic theoretical analysis of the stability and bifurcations of four representative collective states observed in simulations: the desynchronization state, the phase-wave state, the synchronization state, and the mixed state. We show that the macroscopic dynamics are governed by a low-dimensional invariant manifold, namely the generalized Ott–Antonsen manifold, and further establish that the same manifold underlies the collective dynamics of high-dimensional agent models, thereby unifying these seemingly different descriptions within a common reduction framework. Building on this connection, we derive stability conditions and bifurcation boundaries for the collective states and construct a bifurcation diagram in the two-parameter coupling plane, with theoretical predictions validated by large-scale numerical simulations. Our results clarify the dynamical origins of state selection in swarmalator systems and provide a theoretical foundation for analyzing nonequilibrium phase transitions in active-matter systems with coupled internal rhythms and spatial motions.
Read the full abstract

Optimization and Control

Fragment-Free Constraint Programming for the Preemptive Flexible Job Shop Scheduling Problem

This article addresses the preemptive flexible job shop scheduling problem with makespan minimization (FJm | pmtn | C_max). We propose an exact constraint programming approach that completely avoids representing the processing fragments of an operation. Operations are placed in flexible time windows, where Horn's workload condition guarantees feasible preemptive scheduling, and Jackson's preemptive earliest-deadline-first rule reconstructs exact processing segments post-optimization in O(n log n) time per machine. To manage undecided machine assignments, we extend the PreemptiveNoOverlap global constraint to optional intervals within IBM ILOG CP Optimizer, performing overload checking and preemptive edge-finding in O(n log n) time while eliminating infeasible machine alternatives. The CP model is warm-started using a multi-round MILP assignment relaxation with Carlier's head-tail bounds and a window-based greedy heuristic. Extensive computational experiments on 276 pFJSSP benchmark instances and 78 pJSSP instances demonstrate that the proposed method significantly advances the state of the art. Within a 600~s time limit, it improves the upper bounds of the previously published logic-based Benders decomposition by 3.6% on average, increases the number of proven optimal pFJSSP instances from 102 to 180, and closes 74 instances for the first time (including the entire Barnes family). On the pJSSP under identical hardware conditions, it proves 67 of 78 instances optimal (compared to 51 by the baseline solver Mistral) and closes 9 of the 11 long-standing open problems.
Read the full abstract

Robotics

Rhythmic-Semantic Co-Speech Gestures for Emotional Dialogue in Robots

Co-speech gestures can increase engagement and improve speech understanding. Yet, most data-driven robot systems generate either rhythmic beat-like motion or semantically meaningful iconic gestures, but not both. We present an emotion-aware rhythmic-semantic co-speech gesture system that generates iconic and beat behaviors in real time from text and the robot's internal emotional state. We formalize the task of beat-iconic gesture generation using emotion and propose lightweight models that predict iconic timing and intensity, then fill the remaining spans with emotion-conditioned beats at an appropriate intensity level. We train on the BEAT2 dataset and deploy the system on the social robot Haru. In an online user study with 60 participants, our method shows stronger speech-gesture alignment and is preferred over baseline methods.
Read the full abstract

All authors Christian Arzate Cruz, Edwin C. Montiel-Vázquez, Stefanos Gkikas, Miguel Mejia Tobar, Randy Gomez

Programming Languages

Context Types: Bunched Modalities for Unifying Safety and Reachability

Most type systems are designed to verify safety properties: the judgement $e : \tau$ holds if $e$ may evaluate to any value contained in the denotation of $\tau$, but nothing more. More recent type systems have considered the complementary class of reachability properties: in these systems, the typing judgment holds if $e$ must be able to evaluate to every value denoted by $\tau$, and possibly more. This paper introduces a refinement type system for higher-order functional programs where types can capture both safety and reachability properties. Simultaneously supporting both is challenging because each imposes different constraints on how variable bindings in a typing context are interpreted. We address this challenge by treating typing contexts as a kind of resource that mediates which capabilities are available when typing a program. We formalize this notion by developing a resource algebra of contextual capabilities and a bunched modal context logic over this algebra. We then interpret our type system in this logic, establishing soundness with respect to a denotational semantics that unifies safety and reachability guarantees within a single typing judgement.
Read the full abstract

Software Engineering

An Empirical Study on Improving LLM-Based Issue Classification through Domain-Informed Analysis

Issue reports that describe Intrinsic Bugs, Extrinsic Bugs, and Not a Bug cases often use similar language, but they might point to different causes and require different maintenance actions. This makes three-way issue classification difficult for Large Language Models (LLMs), particularly when prompt-based refinements begin to plateau. We study whether domain-informed analysis of LLM outputs can improve classification beyond prompt modification alone. Using 428 manually labeled issue reports, we evaluate six LLMs across 18 prompt configurations and analyze 7,704 model-generated rationales to identify recurring failure patterns in how models justify predictions, use evidence, and handle conflicting information. We translate these findings into a multi-stage pipeline that separates bug validity from bug origin. The pipeline improves macro F1 and Extrinsic Bug recall for five of the six models. The strongest pipeline configuration reaches 0.827 accuracy and 0.757 Extrinsic Bug recall, improving over the best-accuracy prompt-only baseline by +1.2 and +7.5 pp, respectively. The decomposition further reveals that bug origin classification is more reliable than validity, indicating that the bug-validity decision accounts for most of the remaining error. These results show that expert failure analysis can support more reliable LLM-based software engineering classifiers.
Read the full abstract