Number Theory
Enci Wang
Given a
polynomial \(g\) of degree \(\delta>0\) over a finite field, we study monic
polynomials \(f\) of degree \(n\) that can be written as \(f=h+g^k\), where \(h\) is a monic irreducible
polynomial of degree \(n\) and \(k\in\mathbb{N}\) with \(\delta k<n\). Following Erdős, we show that some \(f\) admit at least \(c\log n\) such representations. If \(\delta\rho_g<1\), where \(\rho_g=\varphi_q(g)/|g|\), then a positive proportion of degree-\(n\)
polynomials admits at least two such representations. More generally, for sums of powers \(g^{\lfloor k_i^{r_i}\rfloor}\), we prove that a positive proportion of
polynomials of degree \(n\) is representable whenever \(\sum_{i=1}^{t}r_i^{-1}\ge 1\), while certain residue classes modulo some irreducible
polynomial contain no representable
polynomial.
Read the full abstract ▾
General Relativity and Quantum Cosmology
Probing black holes surrounded by Dust field in Quantum Fluctuation Modified Gravity with twin-peak QPOs
Kamola Shermatova, Qodir Badalov, Abubakir Shermatov et al. (+5 more)
We study a static, spherically symmetric black hole solution surrounded by a
dust field within
quantum fluctuation modified
gravity, in which the
quantum-fluctuation parameter $\alpha$ and the
dust-field strength parameter $K$. The horizon structure reveals the two parameters acting in opposite directions: strengthening $K$ draws the inner and outer horizons together until they merge into a single extremal horizon, while a larger $\alpha$ restores the Schwarzschild limit; the same opposing trend extends to the effective potential and the characteristic circular-orbit radii – the innermost stable circular orbit, the marginally bound orbit, and the marginal circular orbit – which shift inward with $K$ and outward with $\alpha$, and to the Keplerian and radial epicyclic frequencies of massive test particles. From these frequencies we build three twin-peak quasi-periodic oscillation models – relativistic precession, warped-disk, and epicyclic-resonance – along with their $3{:}2$, $4{:}3$, and $5{:}4$ resonance radii, all following the same $K$-$\alpha$ trend. Testing all three models against twin-peak QPO data from two
stellar-mass (GRO~J1655$-$40, XTE~J1550$-$564) and two intermediate-mass (M82~X$-$1, NGC~1313~X$-$1) sources through a Markov Chain Monte Carlo analysis places best-fit constraints, at the $2\sigma$ level, of $K\in(0,0.5)$ and $\alpha\in(0.3,0.6)$ across all sources and models.
Read the full abstract ▾
All authors Kamola Shermatova, Qodir Badalov, Abubakir Shermatov, Javlon Rayimbaev, Bekzod Rahmatov, Shokhzod Jumaniyozov, Dilafruz Xujanova, Farhod Nuraliev
High Energy Astrophysical Phenomena
Dark matter self-annihilation as a hidden energy source in white dwarfs
Alvin Cheuk-Nam Chu, Cheuk-Man Yiu, Ching-Hui Lam et al. (+3 more)
White
dwarfs (WDs) are compact remnants of
stars, supported by electron degeneracy pressure, and their structures are generally considered to be well understood. Significant discrepancies exist between the theoretical and observed white
dwarf mass-radius (MR) relations, especially for hot Q-branch white
dwarfs with unexplained heating sources. In this study, we explore the potential of WDs as a portal to dark matter (DM) physics by investigating the effect of self-annihilating DM (SADM) admixed in WDs, solving the set of two-fluid hydrostatic
equilibrium and heat
diffusion equations self-consistently. We obtain empirical formulae for the WD luminosity $L$ and effective temperature $T_\text{eff}$ as functions of the DM mass fraction $f_d$, annihilation cross section $\langle\sigma v\rangle$, and particle mass $m_\chi$, as well as the WD's total mass $M_t$. We also compare the WD MR relations for various $L$, $f_d$, and both Fermionic and Bosonic DM particle statistics, with the observational data. We show that the majority of observed deviations in the WD MR relation can be accounted for by having SADM admixed with $f_d \le 10^{-5}$ and $\langle\sigma v\rangle \ll 3\times10^{-26}$ $\text{cm}^3$ $\text{s}^{-1}$, the thermal relic cross section. Therefore, our results suggest that WDs can be sensitive DM detectors, and DM self-annihilation could provide additional heating in WDs, accounting for their observed extra luminosities.
Read the full abstract ▾
All authors Alvin Cheuk-Nam Chu, Cheuk-Man Yiu, Ching-Hui Lam, Peter Siu-Hei Cheung, Fong-Ching Ho, Ming-chung Chu
Commutative Algebra
v-numbers of symbolic-ordinary discrepancy modules
Manohar Kumar, Nguyen Cong Minh and Thanh Vu
In this paper, we study the ${\rm v}$-number of the symbolic-ordinary discrepancy modules $M_t(I) = I^{(t)} / I^t$ over standard graded Noetherian rings. We show that this coincides with the ${\rm v}$-number of $I^t$ at embedded
primes. Moreover, under mild assumptions on $I$, we show that the ${\rm v}$-number of $N_t(I) = I^{(t)} / I^{(t+1)}$ is the same as the ${\rm v}$-number of $I^{(t+1)}$. Furthermore, we provide combinatorial formulas for these ${\rm v}$-numbers when $t = 2, 3$ and $I$ is the edge ideal of a graph. Consequently, we classify all graphs for which ${\rm v}(I^2) = {\rm v}(I^{(2)})$.
Read the full abstract ▾
Robotics
A Reconfigurable Dual-Opposition Architecture for Single-Hand Assembly and Manipulation
William Su, Yunosuke Nakamura, Yixiao Wang et al. (+6 more)
In-hand assembly is constrained by the need to maintain grasps on two separate parts while controlling their relative motion within a single hand. To enable both in-hand assembly and manipulation, we present a reconfigurable dual-opposition architecture. Specifically, to support simultaneous grasping of two parts and coordinated in-hand manipulation, four independently actuated fingers are organized into two virtual finger (VF) oppositions, with their relative configuration controlled by a reconfigurable palm. To describe hand motion and simultaneous two-object grasping configurations, a kinematic model of the fingers and palm and an object-size-conditioned workspace formulation are built. To further evaluate motion performance and assembly capability, finger-joint motion and palm tracking are characterized, and in-hand assembly is demonstrated through tasks involving grasping,
alignment, fastening, and pressing. Ablation experiments further demonstrate the importance of finger abduction/adduction and palm reconfiguration for successful in-hand assembly. In simulation, the proposed hand achieves a mean continuous sphere rotation success rate of 98.6% over diameters of 40-230 mm, compared with 73.8% for the LEAP Hand. After policy fine-tuning with external disturbances, the proposed hand achieves 92.8% success under disturbances from multiple directions, compared with 45.2% for the LEAP Hand. Hardware demonstrations further show in-hand rotation of objects of different sizes using policies trained in simulation. Together, these results show that the proposed architecture supports both assembly of two separately held parts and coordinated manipulation of a single object within one hand.
Read the full abstract ▾
All authors William Su, Yunosuke Nakamura, Yixiao Wang, Yitong Li, Mingrui Yu, Huanan Qi, Boyuan Liang, Masayoshi Tomizuka, Jianshu Zhou
Machine Learning
Towards Full Pipeline FP8 Reinforcement Learning for LLMs
Fanchao Chen, Ziheng Jiang, Ziyun Wei et al. (+5 more)
Reinforcement learning (RL) has become a key technique for improving the
reasoning and agentic abilities of large
language models (LLMs). Although FP8 quantization can accelerate RL training, maintaining stability throughout an FP8 RL pipeline remains challenging. While previous works have focused on resolving train-inference mismatches using correction techniques like TIS, we reveal that full-pipeline FP8 RL still suffers from severe training instability, manifesting as anomalous mid-training
entropy surges and garbled outputs. We trace this instability to a previously overlooked cause: compounded FP8 quantization noise distorts the importance ratio, disproportionately pushing negative-advantage
tokens outside the trust region and erroneously zeroing out their gradients. As a result, pathological outputs are not properly penalized and accumulate over the course of training. To address this, we propose Calibrated Clipping, a dynamic method that aligns the FP8 clipping bounds with high-precision BF16 distributions by matching the lower-bound clipping
quantile and rebalancing the upper bound accordingly. Extensive experiments across GRPO and DAPO
algorithms, model scales from 8B to 32B, and multiple FP8 scaling granularities demonstrate that our approach successfully eliminates
entropy surges and restores performance comparable to the BF16 baseline.
Read the full abstract ▾
All authors Fanchao Chen, Ziheng Jiang, Ziyun Wei, Zheng Zhong, Du Li, Chi Zhang, Haibin Lin, Shivaram Venkataraman
Computers and Society
The Moral Check: Strategic AI Governance for the Pacing Problem
Zaid Amin, Rahma Santhi Zinaida and Nazlena Mohamad Ali
Technology cannot steer itself. Strategy provides that steering, establishing the rule that purpose and judgment must precede compute capital. As the frontier artificial intelligence (AI) ecosystem accelerates exponentially, the pacing problem induces severe cognitive tunneling in engineering teams, prioritizing scalar throughput over human judgment. A calibrated pacing rate is imperative to check unchecked scaling, guarantee safety, and build models in whose
alignment society can place warranted confidence. Traditional oversight fails through retrospective checklists, a pathology of performative governance exhibiting high procedural maturity but alarming scientific immaturity. We deliver a dual contribution: a PRISMA 2020 review synthesizing 130 empirical studies (MMAT-appraised across 18
benchmarks; total corpus N = 130 empirical studies across 178 reference foundations), and the Strategic AI Governance Ex-Ante Framework (SAGE-X). Our synthesis exposes two systemic vulnerabilities: the Recursive Assurance Paradox (correlated, ungrounded evaluator confidence) and the Durability Deficit (guardrail decay under multi-turn shifts). Grounded in MIT Strategic Computing doctrines, SAGE-X operationalizes Four Strategic Mindset Pillars: (1) Intent over Execution (mitigating velocity myopia); (2) Ruthless Trade-offs (deterministic tripwires eliminating moral hazard); (3) Outcomes over Outputs (
auditing empirical hazard endpoints); and (4) Proactive
Alignment (synchronizing ex-ante gates with runtime telemetry). Governed by a calculable Moral Check Index (MCI) with an unbypassable tripwire, SAGE-X delivers an operational Enterprise Lifecycle Audit Instrument (the "Moral Check Audit Card") on Stanford WebProtégé, ensuring exponential progress never outpaces deliberative moral judgment, human agency, and societal trust.
Read the full abstract ▾
Computer Vision
Planning-Aligned Pretraining of BEV Representations with Sparse Action-Conditioned Targets for End-to-End Autonomous Driving
Jaeha Song and Soonmin Hwang
End-to-end driving requires planning-relevant bird's-eye-view (BEV) representations, but existing pretraining approaches often rely on task annotations or dense scene reconstruction. We introduce PAVER, Planning-Aligned BEV Encoder Pretraining. From a single
LiDAR sweep, PAVER constructs sparse risk and unknown targets describing occupied and unobserved evidence along rule-based ego motions. A 10K-parameter head predicts these targets from masked BEV features conditioned on the action state, directing supervision toward
geometric constraints on candidate motions. Pretraining requires no driving-task annotations or dense reconstruction. Only the BEV encoder is transferred, preserving the downstream architecture and camera-only inference. On nuScenes, PAVER reduces VAD-Tiny's average collision rate from 0.51% to 0.19%, while improving planning L2, motion prediction, detection, and mapping. The selected VAD-Tiny and VAD-Base schedules use about 36% less estimated total training time than scratch training, including pretraining. On Bench2Drive Town05 Long, PAVER improves UniAD-Tiny's closed-loop Driving Score from 48.45 to 58.79. The project page is available at https://archiiive99.github.io/PAVER.
Read the full abstract ▾
Machine Learning
Leveraging Inference-Time Compute for Diffusion Models via Global Scheduling of Denoising Trajectories
Yuan Cao, Yifu Tang, Hangqi Li et al. (+1 more)
Diffusion models generate a sample by traversing a denoising trajectory, a sequence of
stochastic noise-reduction steps that transforms pure noise into a draw from a target distribution. At deployment time, additional computation can improve sample quality without retraining: at each step, the sampler draws several candidate noise samples, scores the resulting predictions with a quality criterion called the verifier, and retains the best candidate at the cost of one network evaluation per candidate. This raises a resource allocation question: given a fixed budget of function evaluations, how should search effort be distributed across the steps of the denoising trajectory? We formulate this as a computational budget allocation problem. First, we show that, to leading order in the step size, the expected gain from evaluating $K$ candidates at a step factorizes into an endogenous, step-specific sensitivity parameter times a universal sample-size factor equal to the expected best of $K$ standard-normal draws. Second, for a fixed sensitivity profile, the optimal allocation solves a separable concave integer program with water-filling structure; at fixed total sensitivity, its advantage over uniform allocation increases with sensitivity dispersion in the majorization order. Third, we prove that when sensitivities vary across instances, no adaptive policy can avoid worst-case regret that grows linearly in the trajectory length, which motivates a design that anchors the allocation offline and adapts online only to recover instance-specific slack. We extend the analysis from independent random search to a broader family of local search operators, and instantiate it as an implementable
algorithm. Experiments on three families of
diffusion samplers show that the proposed allocation attains the quality of the uniform
benchmark with 20 to 50 percent fewer function evaluations.
Read the full abstract ▾
All authors Yuan Cao, Yifu Tang, Hangqi Li, Zeyu Zheng
Machine Learning
Causilo Technical Report
Minyong Cho, Minho Jeong, Dooho Lee et al. (+2 more)
We introduce Causilo, a tabular foundation model (TFM) that combines frontier predictive performance with exceptionally fast inference. On TabArena, Causilo achieves 1785.4 Elo, at a median inference time of 0.10 seconds per 1K test samples. It outperforms TabPFN-3.5-Fast with 31.6% less inference time, placing it on the performance–efficiency Pareto frontier. Causilo follows TabICL's column-then-row architecture but introduces another row-refinement module before row compression. This module exchanges information among
cell representations within each row after column encoding. The refined
cells then visit the context set again through an additional column stage before being compressed into row
embeddings. For inference efficiency, both row stages use cross-
attention through a fixed number of summary
tokens, keeping their
attention cost linear in the number of features. Pretrained on approximately 36M synthetic tables, Causilo delivers strong
benchmark results across TabArena, BeyondArena, and ScoringBench, achieving frontier-level performance with substantially faster inference.
Read the full abstract ▾
All authors Minyong Cho, Minho Jeong, Dooho Lee, Jinmo Lee, Jaemin Yoo
Systems and Control
AoI-Driven Hierarchical Learning for Cooperative Resource Sharing in Multi-Operator UAV Networks
Atefeh Hajijamali Arani, Mahyar Shirvanimoghaddam, Abolfazl Mehbodniya et al. (+2 more)
Uncrewed aerial vehicle (
UAV)-assisted networks provide a versatile paradigm for on-demand connectivity. However, in multi-operator aerial networks (MOANs), the joint optimization of cooperative resource sharing and 3D trajectory control to maintain information freshness is a complex combinatorial problem, which can be shown to be NP-hard. To address this computational complexity, we propose an age of information (AoI)-driven hierarchical deep reinforcement learning (DRL) framework. Specifically, a Dueling Double Deep Q-Network (D3QN) architecture is deployed at both the operator and
UAV decision layers to mitigate overestimation bias and enhance stability in high-dimensional state spaces. To improve system resilience, we introduce an AoI- and load-aware outage compensation mechanism that prioritizes users based on instantaneous transmission demands and temporal freshness. Furthermore, a normalized load exchange balance metric is incorporated to regulate cooperative behavior and ensure resource
fairness across operators. Simulation results demonstrate that the proposed hierarchical D3QN significantly outperforms conventional DRL, non-cooperative, and cooperative
benchmarks, reducing the average AoI by up to 56.1% under severe congestion while ensuring superior inter-operator
fairness and outage mitigation.
Read the full abstract ▾
All authors Atefeh Hajijamali Arani, Mahyar Shirvanimoghaddam, Abolfazl Mehbodniya, Halim Yanikomeroglu, Fumiyuki Adachi
Numerical Analysis
Asymptotic-Preserving Exponential Integrators applied to the Hyperbolic Korteweg-de Vries System
Abhijit Biswas, Sebastiano Boscarino, David I. Ketcheson et al. (+1 more)
We study the application of exponential time integration methods, coupled with
Fourier pseudospectral space discretization, to the
hyperbolic Korteweg-de Vries (KdVH) system. We investigate the asymptotic preserving (AP) properties of such discretizations, showing that, in general, Lawson methods are not asymptotic preserving, because the auxiliary (derivative-approximating) variables do not satisfy the limit
equilibrium manifold, whereas exponential time differencing (ETD) methods are AP for all components, including both the solution variable and the auxiliary variables. We also present an efficient numerical implementation based on an exact formula for the
matrix exponential of this system, and compare it with previously-proposed ImEx Runge-Kutta time integration, showing that exponential methods can be competitive in this context.
Read the full abstract ▾
All authors Abhijit Biswas, Sebastiano Boscarino, David I. Ketcheson, Giovanni Russo
Physics and Society
Below the Surface: Creep, Corrosion, and Tipping Points in the Social Resilience of Science
Pawel Sobkowicz
By its most visible quantitative metrics: numbers of researchers, papers, journals, and institutions, science has never looked healthier. Yet a set of subsurface indicators points the other way: the disruptiveness of the average paper and patent is falling, progress slows in the largest fields, replication is rare and often unsuccessful. Incentive systems can select for poor methods even when no one is actually cheating. Lay society shows a curious mixture of admiration and lack of trust for scientists and their work. We argue that a healthy-looking surface can coexist with, and mask, an accumulation of below-the-surface damage due to slow-acting, internally and externally driven pressures, weakening the social position of science. This process is best understood not as an acute shock but as a creeping crisis. And while the damage grows slowly, it may result in a sudden collapse of the research as social institution. To make this diagnosis tractable rather than merely alarming, we import the conceptual and mathematical apparatus of two mature research fields that already study such slow degradation and sudden collapse processes: the continuum damage mechanics of creep, fatigue, and corrosion, and the ecology of resilience, regime shifts, and critical transitions. We do not claim to resolve whether science is merely robust under strain or is approaching a tipping point; the available data are too fragmentary. This remains a question of importance and we propose what would answer it: finding measurable early-warning indicators, use of
agent-based models of the coupled trust-incentive-funding feedback system, understanding
adversarial attacks, and the design of resilience building and recovery interventions. The result is a research agenda, and an argument for why the health of science cannot be read off its own success reports.
Read the full abstract ▾
Machine Learning
CurvFlow-DTA: dual-graph discrete Ricci curvature flow for drug–target affinity prediction
Jicheng Ma, Yunyan Yang, Juan Zhao et al. (+1 more)
Graph neural networks are widely used for drug–target affinity (DTA) prediction, and discrete
Ricci curvature has recently been used to characterize molecular graph geometry. Existing
curvature-aware DTA approaches mainly use static
curvature on the drug graph while representing
proteins primarily with sequence-derived features. This leaves pair-adaptive use of graph geometry underexplored, which may limit adaptation to unseen entities in cold-start settings relevant to practical screening. We present CurvFlow-DTA, which replaces a single static
curvature representation with weighted Forman
curvature flow on both molecular and
protein residue–residue contact graphs. A label-independent flow trajectory is precomputed for each entity, and a pair-conditioned selector determines the horizons read by a dual-branch Flow-GINE. A frozen ESM-2 supplies residue-level representations and contact scores used to construct the
protein graph. Inference requires only SMILES strings and
protein sequences, without a bound complex structure. On Davis and KIBA, CurvFlow-DTA improves on the protocol-matched
Ricci-GraphDTA baseline in every warm and cold-start setting. Warm-split mean squared error (MSE) decreases by $19.9\%$ on Davis and $18.9\%$ on KIBA. Across the six cold-start comparisons, MSE decreases by $14.3$–$27.4\%$, with higher concordance index (CI) throughout. Within our compiled set of literature baselines, CurvFlow-DTA achieves the lowest MSE on both warm
benchmarks and across four out of six cold-start evaluation settings.
Read the full abstract ▾
All authors Jicheng Ma, Yunyan Yang, Juan Zhao, Liang Zhao
Mathematical Physics
Holomorphic Conformal Cartan Geometry: Conformal Gravity, Twistors, and Ambitwistors
Chen-Hsu Chien
Four-dimensional conformal
gravity, twistor theory, and ambitwistor theory are reviewed within the unified framework of holomorphic conformal Cartan geometry. Starting from the spinorial representation of the geometry, the normal $\mathfrak{sl}(4,\mathbb{C})$-valued Cartan connection, its
curvature $2$-form, and the associated tractor bundles are systematically formulated. Evaluated on the real Lorentzian slice, the quadratic
curvature action is shown to reproduce four-dimensional Weyl–Bach conformal
gravity upon fixing a Weyl gauge. Local twistor and dual twistor bundles are constructed as associated
spin tractor bundles, illustrating how parallel transport yields the classical incidence relations in flat space and
chiral integrability constraints in curved backgrounds: anti-self-duality for $\alpha$-surfaces and self-duality for $\beta$-surfaces. Furthermore, by pairing local twistors and dual twistors via an invariant canonical pairing, ambitwistor space is shown to parameterize complex null geodesics and symmetrically accommodate both left- and right-handed Weyl
curvatures. This framework establishes a direct gauge-theoretic bridge linking conformal Cartan geometry to twistor and ambitwistor theories.
Read the full abstract ▾
General Relativity and Quantum Cosmology
Charged Klein-Gordon modes on an Ellis wormhole: magnetic confinement and conditional Heun solvability
Abdullah Guvendi and Omar Mustafa
We study charged Klein-Gordon modes on a $(2+1)$-dimensional Ellis traversable wormhole in an external
magnetic field defined as uniform in an auxiliary Euclidean
embedding space. The spatial geometry is intrinsically catenoidal, while the
embedding is used only to construct the external gauge connection. Pullback of the ambient potential yields a regular azimuthal gauge field whose associated intrinsic
magnetic scalar reverses sign across the throat. Separation of the Klein-Gordon equation leads to a radial problem that is unitarily equivalent to a one-dimensional Schrödinger operator with a regular effective potential. For nonzero
magnetic coupling, the quadratic asymptotic term produces radial confinement on the complete wormhole. The radial equation can be reduced to the confluent-Heun class.
Polynomial truncation is possible only on constrained parameter submanifolds determined by simultaneous termination conditions and therefore does not represent the generic confined spectrum. We derive the corresponding relativistic energies and analyze the lowest
polynomial sector, while distinguishing the formal continuous-$\ell$ solvability correlations from the integer values of the azimuthal separation constant required by the standard angular periodicity condition.
Read the full abstract ▾
Astrophysics of Galaxies
The interplay between active galactic nucleus photoionization, radio jet, and star formation in the z $\sim$ 3.5 radio galaxy 4C +03.24
Bruno Dall'Agnol de Oliveira, Dominika Wylezalek, Pranav Kukreti et al. (+9 more)
High-
redshift radio galaxies (HzRGs) are among the most powerful
radio sources, and are associated with the most massive
galaxies and dense environments at
redshifts z $\gtrsim$ 1. They are ideal laboratories for studying how active
galactic nucleus (AGN) events can shape the
evolution of
galaxies, as intense radiation, jets, and
star formation can be observed simultaneously in these
galaxies. We present JWST/NIRSpec integral field spectroscopy ($\sim$ 1.6 kpc spatial resolution) of the 4C +03.24 system, a powerful HzRG at z $\sim$ 3.5 with a bolometric luminosity of $\sim 10^{47.6}$ erg s$^{-1}$. We identified kinematically disturbed regions in the warm ($\sim 10^4$ K) ionized gas by decomposing the emission-line spectra into multiple Gaussian components, which is crucial to avoid overestimating the
outflow properties. The
outflow power peaks at $\sim$ 2 kpc away from the
nucleus, with a corresponding low kinetic coupling efficiency of $\sim 8_{-5}^{+7} \times 10^{-3}$ %. A combined analysis of the rest-frame optical and ultraviolet (from VLT/MUSE and HST imaging) continua revealed an extended emission (spanning $\sim$ 14 kpc), which we interpret as partially tracing
star-forming regions. With a clearly delineated bipolar morphology, we show that the AGN photoionization dominates the ionization of the interstellar medium along the
radio jet axis. The [C II]$\lambda$158$\mu$m emission gap in this region might be direct evidence of negative AGN feedback. We also discuss a possible scenario where 4C +03.24 could be situated in an overdense environment experiencing multiple
galaxy interactions and the possibility of jet-induced
star-formation.
Read the full abstract ▾
All authors Bruno Dall'Agnol de Oliveira, Dominika Wylezalek, Pranav Kukreti, Wuji Wang, Rogério Riffel, Rogemar A. Riffel, Andrey Vayner, Caroline Bertemes, Julian T. Groth, David S. N. Rupke, Carlos De Breuck, Joël Vernet
Robotics
PileBelief: Persistent Physical State for Interaction-Driven World Modeling
Hongyi Lin, Song Zhang, Haiquan Liu et al. (+3 more)
World models allow
robots to anticipate action consequences before execution. This capability is especially valuable in excavation, where each scoop reshapes the terrain and affects subsequent actions. Local observations, however, cannot fully reveal the underlying support and material conditions. We present PileBelief, an interaction-driven persistent world model for partially observed excavation that retains physical evidence beyond the visible surface. It combines an observation-conditioned physical prior with world-addressed deformation
memory and physical-response
memory. Action-aligned reads and gated residual corrections refine terrain-change and outcome predictions. With deployment weights fixed, completed interactions update measured belief, while hypothetical actions advance a separate imagined state. Compared with a current-observation-only baseline, PileBelief reduces five-step joint prediction error by 10.8% and offline action-selection regret by 65.5%. Experiments on Newton/MPM and real excavation
datasets further demonstrate improved terrain-change and bucket-volume prediction. Our method enables multi-step prediction and candidate-action ranking from local observations, even when the underlying soil state is unknown. These results identify persistent physical belief as a useful representation for world models of environments that
robots continually reshape.
Read the full abstract ▾
All authors Hongyi Lin, Song Zhang, Haiquan Liu, Yang Liu, Jinhua Zhao, Xiaobo Qu
Computer Vision
Image Frame Dynamic Object Segmentation and Ego Motion Estimation using Radar Image Fusion
Astik Srivastava, Suhani Grover, Avinash Sharma et al. (+1 more)
Dynamic object
segmentation and ego-motion estimation are closely coupled problems in autonomous driving, as accurate ego-motion estimation typically requires static scene observations, while identifying static observations requires knowledge of the ego motion. We present Radar-Dot, a radar–RGB framework that exploits radar Doppler measurements to address this coupling. Radar returns are first used to estimate ego velocity through a linear Doppler constraint, with residual-based static/dynamic
segmentation and robust estimation used to reduce the influence of moving objects. The estimated motion is then combined with metric depth and dense optical flow to identify image regions whose observed motion is inconsistent with the rigid scene motion. Experiments on 10 nuScenes scenes (part of nuscenes-mini) demonstrate that the resulting
geometric pipeline achieves 20.24% dynamic IoU and 33.67% F1-score over 394 frame pairs, while radar-based static-point filtering improves ego-velocity estimation compared with using all radar returns. These results demonstrate the potential of radar as a modality for jointly improving ego-motion estimation and dynamic object
segmentation.
Read the full abstract ▾
All authors Astik Srivastava, Suhani Grover, Avinash Sharma, Madhava Krishna
Image and Video Processing
A Controlled Evaluation of Quantum Correlation Refinement for Few-Shot Semantic Segmentation: Resource Cost and IBM Quantum Hardware Validation
Hina Shakir, Muhammad Irfan Memon, Asia Samreen et al. (+4 more)
PQCs are increasingly proposed as trainable components in classical ML pipelines, but rarely characterized alongside a controlled measurement of task-level benefit. We report such an evaluation, using few shot
segmentation as testbed. We integrate a
Quantum Correlation Refiner (QCR), a six-
qubit variational
circuit with amplitude
embedding and a gated residual connection into a classical correlation-based architecture, under the four fold PASCAL 5i protocol, three seeds per fold, against an unrefined baseline and a matched MLP refiner. Across 12 fold-seed comparisons, QCR changes mIoU by only +0.0001 (t(11)=0.10, p=0.92, dz=0.03); the MLP control and an expanded Pauli-measurement variant show similarly no improvement. QCR adds just 473 parameters but roughly doubles per epoch training time. A trained
circuit on IBM hardware (18 patches) agrees closely with the noiseless simulator (r=0.954, MAE=0.101), a fidelity check, not an accuracy gain. The module is functional, trainable, and hardware deployable, yet shows no task-level advantage over a matched classical alternative under this regime, a template for evaluating
quantum components jointly via matched controls, repeated seeds, resource measurement, and hardware validation, rather than any one alone.
Read the full abstract ▾
All authors Hina Shakir, Muhammad Irfan Memon, Asia Samreen, Javeed Hussain, Syed Rizwan Ali, Mohammad Mohatram, Muhammad Hussain
Probability
Asymptotic estimations for a nonstandard bidimensional risk model with dependent claims and constant interest force
Dimitrios G. Konstantinides and Charalampos D. Passalidis
In this paper we extend the results from Yang and Li(2017), in a non-renewal driven bivariate risk model. Concretely, we are interested in the asymptotic behavior of the joint tail of the discounted aggregate claims over finite and infinite time horizon in a bivariate risk model with two arbitrarily dependent counting processes. We additionally suppose that the sequences of claim-sizes of the two lines of business are independent, but each of them contains weakly dependent terms. In our main results, on finite and on infinite time horizon, we assume two (different) general conditions for the counting processes, that are satisfied by a wide spectrum of processes beyond the renewal ones. In the case of finite horizon we suppose that the claim distributions from the two lines, belong to the subexponential distribution class, while in the case of infinite horizon we restrict ourselves to the consistently varying and positively decreasing distribution class. More explicit asymptotic expressions are derived in the case we are restricted on the regularly varying class for the claim distributions. We note that our results indicate the presence of multivariate non-linear single big jump principle for the discounted aggregate claims.
Read the full abstract ▾
Robotics
SmoLSTM: A Compact Vision-Language-Action Model with Recurrent Memory that Persists
Jan-Gerrit Habekost, Parsa Mastouri Kashani, Connor Gäde et al. (+5 more)
Vision-
language-action models often predict actions from only the current observation, which can leave tasks involving object occlusion or visually identical objects ambiguous without episode history. The usual countermeasure, widening the observation window, turns the horizon into a hyperparameter and lets per-step cost grow with it. We instead capture the episode in a recurrent state. SmoLSTM couples a frozen 256M-parameter SmolVLM backbone to a
matrix-
memory LSTM control layer in which observation
tokens and action queries are unified in a single causal stream that is never reset throughout the entire episode. Recurrent-state storage is therefore O(1) in episode length. A flow-matching action head predicts chunks of 10 end-effector pose deltas and gripper commands at each control step. Our single policy, trained jointly on 7,461 demonstrations across 140 tasks and evaluated on held-out initial states, performs best, reaching 85.1% subgoal coverage and 77.5% full-task success on LIBERO-Mem with 0.04B trainable parameters, surpassing both the
benchmark's own object-centric baseline and recent
memory-based approaches. Resetting the recurrent state at every control step reduces full-task success to 7.0%, showing that the trained policy relies on context carried between decisions. The same model achieves 79.6% average success on standard LIBERO.
Read the full abstract ▾
All authors Jan-Gerrit Habekost, Parsa Mastouri Kashani, Connor Gäde, Matthias Kerzel, Philipp Allgeuer, Cornelius Weber, Stefan Wermter, Jae Hee Lee
Commutative Algebra
On the n-perfectness under split nilpotent extensions
Wei Qi and Xiaolei Zhang
Let $\pi:S\twoheadrightarrow R$ be a split ring epimorphism with nilpotent kernel $J$, and let $R\hookrightarrow S$ be a fixed ring-theoretic section. We prove a change-of-rings formula which is considerably stronger than an equality of global invariants: if $X$ is a left $S$-module satisfying $\Tor_i^S(R,X)=0$ for every $i>0$, then \[ \pd_S X=\pd_R(R\otimes_S X) \] in $\mathbb N\cup\{0,\infty\}$. In particular, for every flat left $S$-module $F$, \[ \pd_SF=\pd_R(F/JF), \] and $F/JF$ is flat over $R$. Reduction and induction of scalars therefore show that the same numerical set of projective dimensions is attained by flat modules over $S$ and over $R$. Consequently, left $n$-perfectness is invariant under every split-by-nilpotent extension. The right-handed analogue holds as well. For a commutative ring $R$ and an arbitrary $R$-module $M$, this applies to the idealization $S=R\ltimes M$ without any projectivity, flatness, finite generation, or finite flat-dimension assumption on $M$. We obtain a list of equivalent characterizations of the condition that $R\ltimes M$ is $n$-perfect, together with \[ \cotD(R\ltimes M)=\cotD(R). \] We further derive exact preservation of the flat-projective-dimension spectrum, strict $n$-perfectness, pointwise localization, arbitrary iterated idealizations, $n$-trivial extensions, truncated
polynomial rings, and upper triangular
matrix rings.
Read the full abstract ▾
Robotics
Physical-Touch Observability from Wrist Wrench in Granular Scooping
Hongyi Lin, Song Zhang, Xubo Liu et al. (+1 more)
Mining and earthmoving are important real-world deployment settings for
embodied intelligence. Autonomous transport and driving systems have improved substantially, but loading and scooping still often depend on skilled human operators, exposing personnel and equipment to operational risk. For robotic scooping, pre-contact RGB-D sensing reveals surface geometry but not the resistance, compaction, tool engagement, or load transfer that emerge during interaction. We test whether the current scoop's six-axis wrist force/torque (F/T), or wrist wrench, contains information about final collected volume, and whether that information depends on the correctly paired action-terrain interaction. We call this property physical-touch observability. Using 6,700 real-
robot scoops across 67 terrains, we evaluate correctly paired current-scoop F/T against pre-contact prediction and correspondence-breaking controls under terrain-held-out testing. At the retrospective 60% sequence boundary, correctly paired F/T reduces mean absolute error by 14.2% relative to Action-only and by 21.9% relative to cross-terrain mismatched F/T. Engineered signal summaries reproduce the result across model architectures. Together, these findings position wrist wrench not merely as a low-level feedback signal, but as a task-level perceptual modality through which
embodied robots can infer hidden physical states during interaction, providing a foundation for response-aware autonomy in mining and other contact-rich tasks.
Read the full abstract ▾
All authors Hongyi Lin, Song Zhang, Xubo Liu, Yang Liu
Sound
Discrete vs. Continuous: A Comprehensive Study of Unified Audio Understanding in LALMs
Jing Peng, Zichao Nie, Zhisheng Zhang et al. (+2 more)
Large Audio
Language Models (LALMs) utilize either continuous features or discrete
tokens, yet the optimal representation paradigm for general audio understanding remains debated. Existing
benchmarks often focus on narrow domains or evaluate encoders outside LALM contexts. To address these gaps, we systematically evaluate continuous and discrete representations across
speech, sound and music. Utilizing our UniARC framework with dual evaluation strategies across model scales from SmolLM2-135M to Llama-3-8B, we analyze the dynamic relationships of data volume, model capacity, and computational efficiency. Our results reveal the pivotal role of semantic constraints in tokenization for audio understanding and demonstrate that scaling backbones fail to compensate for information loss in audio representation, especially in data-limited tasks. These findings offer practical guidance for balancing semantic density, fidelity, and efficiency in future LALMs.
Read the full abstract ▾
All authors Jing Peng, Zichao Nie, Zhisheng Zhang, Jingran Xie, Zhiyong Wu
Machine Learning
Same Outcome, Different Readout: What Does a Steerable Valence Direction in LLMs Represent?
Weihan Li, Xinlei Chen, Yuhan Song et al. (+2 more)
Claims about what an internal direction in an LLM represents need evidential constraints beyond an observer's prior beliefs about the system. Decodability and successful activation steering do not, by themselves, establish which construct the direction tracks. This gap is especially consequential for welfare-relevant interpretations, where a proposed functional state must be distinguished from correlated features of the extraction contrast. We treat the question as one of construct validity and study a good-bad outcome direction in a maze task, using controlled interventions that separate the realised outcome from the informational history through which it became known. Across multiple LLM checkpoints, directions fitted on one explicit outcome encoding transfer well to another, indicating that the readout is not tied to surface form. When the same realised outcome is reached through announced and unannounced histories, however, transfer degrades substantially: even after both histories receive the same explicit outcome, the post-event readout remains strongly conditioned on the earlier announcement. In the base model, steering along the direction changes actions, yet removing it leaves the natural cue effect almost intact. In a matched maze-RL run, the post-RL direction becomes substantially more predictive of reference-MDP remaining return and the policy becomes more dependent on it at the tested sites, while the history dependence persists. These dissociations support a functional, value-related interpretation of the direction, but not its identification with a history-invariant scalar valence state.
Read the full abstract ▾
All authors Weihan Li, Xinlei Chen, Yuhan Song, Xiaofeng Lin, Tianshi Zheng
Computer Vision
LINGO: Latent Initialization and Gradient Optimization for Sparse-view X-ray Novel View Synthesis and CT Reconstruction with 3D Gaussian Splatting
Lifeng Xing, Dequan Jin, Kunpeng Bu et al. (+2 more)
In novel view synthesis and Computed
Tomography (CT) reconstruction with sparse-view X-ray imaging, insufficient angular coverage leads to structural ambiguity and accumulated noise. Integrating 3D Gaussian Splatting (3DGS) with X-ray absorption physics can achieve promising results, but it suffers from noisy initialization, positional insensitivity, and weak gradients in low-density regions. In this paper, we propose a unified Latent Initialization and Gradient Optimization (LINGO) framework to address these issues. LINGO combines latent mask-space initialization with dynamic gradient optimization to improve point cloud structural completeness while accelerating training. It constructs voxel-level 3D filters from X-ray masks to robustly suppress background noise and provide reliable
geometric priors. By employing an adaptive voxel scaling strategy and dynamically scaling loss, LINGO can adjust spatial resolution and explicitly amplify gradients in low-density structures. To evaluate the quality of initialization, we introduce the Initialization Point Cloud Structural Deviation (IPSD) metric. Experiments on the X3D
dataset indicate that for the novel view synthesis task, LINGO improves the Peak Signal-to-Noise Ratio (PSNR) and Structural Similarity Index (SSIM) by an average of 0.72 and 0.0039, respectively, over baselines under identical sparse-view settings, achieving comparable reconstruction quality within 5k steps to state-of-the-art models typically trained with 30k iterations. For the CT reconstruction task, LINGO also demonstrates consistent improvements, with average PSNR and SSIM gains of 0.36 and 0.0134. These results highlight LINGO's effectiveness in both accelerating training and enhancing reconstruction quality across different sparse-view imaging scenarios.
Read the full abstract ▾
All authors Lifeng Xing, Dequan Jin, Kunpeng Bu, Peigeng He, Shihui Ying
Networking
RoutingBench: Can Agentic Routing Analysis Scale to Production Datacenter Networks?
Wenlong Ding, Zhixiong Niu, Jianan Yang et al. (+6 more)
Recent advances in AI models and agentic technologies make AI for network operations (NetOps) within reach. However, scalability remains a key bottleneck of agentic NetOps when analyzing hyperscale networks, which comprise hundreds of datacenters, each housing thousands of network devices. The scalability challenge is rooted in the requirement of many NetOps tasks that must conduct global
reasoning on how a local change of device behavior affects all relevant routing paths, known as routing-path analysis. This paper studies this scalability problem and evaluates how different agentic approaches, namely in-context learning, iterative
reasoning, and
agent skills, can scale routing-path analysis to large, complex networks. We present RoutingBench for evaluating agentic routing-path analysis, with varying network size and complexity, for various types of device changes. Our results show that agentic analysis is promising:
agent skills curated with a principle termed "explore more; digest less" enables routing-path analysis on hyperscale networks of 50K routers with an accuracy of 99.5%, significantly outreaching the scalability of traditional symbolic analysis. Meanwhile, RoutingBench also reveals the boundary of AI
agent capability on complex inter-datacenter networks and compound changes, posing open challenges for AI and agentic research.
Read the full abstract ▾
All authors Wenlong Ding, Zhixiong Niu, Jianan Yang, Fajun Zhang, Bo Zhang, Ling Liang, Yongqiang Xiong, Tianyin Xu, Hong Xu
Rings and Algebras
$\boldsymbol{i}$-conjugate for quaternionic matrices and related properties
Cailing Yao, Bingzhe Hou and Xiaoqi Feng
Motivated by the result that a complex $n\times n$
matrix $A$ being unitarily equivalent to a real
matrix, we extend the conclusion to the quaternion skew field in this paper, we present a necessary and sufficient condition for that a quaternion $n\times n$
matrix $A$ is unitarily equivalent to a complex
matrix. To state the truth more clearly, we put forward the concept which we call $\boldsymbol{i}$-conjugate. Furthermore, we study the concepts related to $\boldsymbol{i}$-conjugate and their properties, such as
unitary $\boldsymbol{i}$-congruence, $\boldsymbol{i}$-conjugate normality and $\boldsymbol{i}$-Hermicity in $M_{n}(\mathbb{H})$ as generalizations of the conventional
unitary congruence, conjugate normality and Hermicity of
matrices in $M_{n}(\mathbb{C})$. Finally, we present a new type of polar decomoposition of quaternion
matrices.
Read the full abstract ▾
Astrophysics of Galaxies
Mid-Infrared Diagnostics of PAH Processing and Radiation Fields in M33 H II Regions with JWST/MIRI
Puja Porel, Archana Soam and Saikhom Pravash
We present JWST/MIRI (Proposal ID 4297) medium-resolution integral field spectroscopy of five H II regions in the nearby spiral
galaxy M33, obtained from the Mikulski Archive for Space
Telescopes (MAST). The fully calibrated Level-3 spectral cubes provide continuous wavelength coverage from 6.5 - 20.9~$\mu$m, enabling analysis of polycyclic aromatic hydrocarbon (PAH) emission, ionic fine-structure lines, and warm molecular hydrogen. The spectra show prominent PAH bands at 7.7, 8.6, and 11.3~$\mu$m, multiple H$_2$ rotational transitions, and hydrogen recombination lines. H$_2$ rotational diagrams yield warm molecular gas temperatures of $\sim316$ - $407$~K. The detected [Ne II], [Ne III], [S III], and [S IV] lines yield ionization indices of $-1.41$ to $+0.98$. The PAH ionization fraction shows no clear dependence on radiation hardness, consistent with previous Spitzer results. Unlike previous work, we find no clear correlation between the 7.7/11.2 and 8.6/11.2~$\mu$m PAH ratios, while the 7.7/11.2~$\mu$m ratio shows a moderate negative correlation with
metallicity, possibly due to our small sample ($N=5$). The PAH-to-VSG ratio decreases with increasing radiation hardness. We derive the ionization parameter $\gamma$, far-ultraviolet radiation field, and PAH charge state, finding a predominantly cationic PAH population ($\sim75%$) despite [Ne III]/[Ne II] $<1$. The lack of a one-to-one correspondence between $\gamma$ and the 7.7/11.2~$\mu$m ratio suggests that PAH ionization is influenced by additional environmental factors. Given the limited sample, these trends are suggestive rather than statistically definitive.
Read the full abstract ▾