Astrophysics of Galaxies
A Full-Sky SPHEREx Map of Integrated 3.3 and 3.4 {\mu}m PAH Emission
Roke Cepeda-Arroita, Clive Dickinson, J. Alberto Rubiño-Martín et al. (+4 more)
Polycyclic aromatic hydrocarbons (PAHs) are ubiquitous in the interstellar medium and are among its smallest carbonaceous grains. Their infrared emission bands trace this small-grain population together with the interstellar radiation field that excites it, and are widely used to diagnose photodissociation regions,
dust processing and
star formation. We present full-sky maps of the integrated radiance of the 3.3 $\mu$m aromatic C-H stretch and the adjacent 3.4-3.5 $\mu$m aliphatic complex, from SPHEREx QR2 Level-2 data. Existing all-sky PAH tracers are broadband, so the band emission can only be inferred as an excess over
stellar and continuum emission; SPHEREx samples the spectrum finely enough for the bands to be fitted directly. We mask flagged detector pixels, subtract the per-pixel zodiacal-light model, subtract point sources with an empirically derived PSF, and coadd the results into inverse-variance-weighted HEALPix maps at 2.6' FWHM and Nside=4096 in 102 narrowband channels across Detector 4 (2.42-3.82 $\mu$m). At every sky pixel we fit a continuum and the two bands. The combined band radiance is the primary product; its decomposition into the aromatic and aliphatic components is released alongside it as an estimate, with its modelling systematic stated. These are the first full-sky maps of PAH band emission, resolving the 3.3 and 3.4-3.5 $\mu$m bands spectrally, and the first full-sky PAH maps from SPHEREx to be released. Because the 3.3 $\mu$m band is emitted preferentially by the smallest PAHs, the maps trace directly a grain population, including the candidate carriers of anomalous microwave emission, that broadband PAH proxies reach only indirectly. The maps and their associated products, including a reliability mask, are publicly available with an explanatory supplement at https://research.iac.es/proyecto/radioforegroundsplus/pages/data-products/spherex-pahs.php.
Read the full abstract ▾
All authors Roke Cepeda-Arroita, Clive Dickinson, J. Alberto Rubiño-Martín, Ricardo T. Génova-Santos, Gabriel A. Hoerning, Stuart E. Harper, Debabrata Adak
Combinatorics
Improved bounds on completion of partial Latin squares
Jack Allsop, Candida Bowtell, Thomas Lesgourgues et al. (+1 more)
A Latin square of order $n$ is an $n \times n$ array filled with $n$ symbols so that each symbol appears exactly once in every row and column. A partial Latin square of order $n$ is an $n \times n$ array whose
cells are either empty or filled in such a way that each symbol appears at most once in every row and column, and at most $n$ distinct symbols are used. In 1983, Daykin and Häggkvist conjectured that every partial Latin square in which each row and column contains at most $n/4$ symbols, and each symbol is used at most $n/4$ times, can be completed to a Latin square. We prove that every partial Latin square in which each row and column contains at most $0.231n$ symbols, and each symbol is used at most $0.231n$ times, can be completed to a Latin square, significantly improving the previous best-known bound of $0.08n$, obtained by Fu and Weng. This problem can be seen as a partite analogue of the Nash-Williams
conjecture concerning triangle decompositions of dense graphs, recently proved in the breakthrough work of Delcourt and Postle. Our proof uses a `discharging' strategy, adapting the approach of Delcourt and Postle, combined with a novel method to achieve a `balancedness' property, required for the partite setting.
Read the full abstract ▾
All authors Jack Allsop, Candida Bowtell, Thomas Lesgourgues, Kalina Petrova
Combinatorics
The topological Bárány-Larman conjecture for prime numbers
Pablo Soberón
The Bárány-Larman
conjecture states that for any $d+1$ sets of $r$ points each in $\mathbb{R}^d$, considered as color classes, we can partition their union into $r$ rainbow $(d+1)$-tuples whose convex hulls intersect. We prove that the topological version of this
conjecture holds when $r$ is a
prime number. We also show that the optimal colorful Tverberg
theorem of Blagojević, Matschke, and Ziegler cannot be extended to
prime powers.
Read the full abstract ▾
Artificial Intelligence
Co-PiLOT: Constrained Physics-Informed Latent Optimization for Target-Driven Inverse Design
Mahish K. Guru, Mayank Nagar, Ayush vyas et al. (+3 more)
Inverse design of physical systems (molecules, devices, microstructures) often reduces to optimizing a high-dimensional structure against an expensive black-box simulator. Direct search is difficult because the space is non-Euclidean, feasibility is hard to encode, and each evaluation is expensive. We present Co-PiLOT, a latent optimization approach that maps candidates through a generative encoder-decoder, uses the decoder as a learned validity prior, and searches the latent space with physics-informed black-box optimization. The framework is applied on the inverse design of magnesium
alloy microstructure/texture. We develop a
vision transformer based-encoder; paired with latent
diffusion,
diffusion transformer and rectified-flow
transformer-based decoders on $\sim80{,}000$ EBSD-derived microstructure
dataset to learn a minimal bottleneck, $z$. The ViT-FMDiT model ($z$=$768$) reconstructs high-fidelity microstructure images (FID $27.86$, MS-SSIM $0.178$), which our self-segmenting orientation codec converts into input grids for
crystal plasticity solver. Finally, we introduce MERIDIAN, an active latent optimizer driven by deep-kernel Gaussian-process uncertainty, failure-aware feasibility prediction,
manifold-aware trust regions, and target-aware acquisition. Within a budget of $160$ simulations, the ViT-FMDiT and MERIDIAN combination yields the best target-driven objective score, reducing the relative target error by $3$–$22\%$ against seven baselines (DANTE, TuRBO, BAxUS, CMA-ES, DDOM, SEIKO, DDPO) on the same decoder.
Read the full abstract ▾
All authors Mahish K. Guru, Mayank Nagar, Ayush vyas, Jan Bohlen, Roland Aydin, Noomane Ben Khalifa
Computer Vision
EndoPrior-GS: Dynamic Endoscopic Reconstruction with a Joint Texture Prior
Jiaqi Huang, Shidong Wang, Tong Xin et al. (+1 more)
Dynamic endoscopic reconstruction is fundamental to robotic surgery and computer-assisted interventions. While 3D Gaussian Splatting (3DGS) realises real-time rendering, its application to deformable intraoperative environments remains constrained by spurious geometry and varying illuminations. To address these limitations, we introduce EndoPrior-GS, a novel pipeline that explicitly couples frame-extracted
vision heuristics and estimated depth maps. EndoPrior-GS derives a joint texture prior from a tool-filtered valid tissue mask, a non-specular photometric filter, and anatomical structural salience, yielding a probability map that guides primitive initialisation and subsequent density control. The prior is further extended to the temporal domain through a texture-aware term that dynamically weighs pairwise primitive contributions during training. We conduct extensive experiments on
benchmark datasets EndoNeRF and SCARED, and the obtained results show that our method EndoPrior-GS reduces Flow Error by 27.7% and 25.8% over the representative approaches while preserving competitive rendering quality and real-time rendering speed. Our project website is available at https://jiaqi-huang-77.github.io/EndoPrior-GS/.
Read the full abstract ▾
All authors Jiaqi Huang, Shidong Wang, Tong Xin, Kabita Adhikari
Statistical Methodology
Target-Aware Sequential Inference: Pooled versus Stratified Anytime-Valid Designs
Subir Hait
Sequential studies with heterogeneous strata may define the estimand under a target mixture that differs from the sampling distribution. We organize the design problem by guarantee class: one pooled confidence sequence for a single prespecified target, shared pooled inference for a prespecified target set, and stratum-resolved inference when local guarantees must remain available. For target-set pooled inference we derive minimax, gap-aware, and range-capped shared proposals and an adaptive tracking rule. A specialized augmented off-policy score removes between-stratum mean heterogeneity asymptotically, recovering Neyman allocation for efficient single-target pooled inference. Robust directional certification over a convex target set is a classical intersection-union problem and does not require an M-fold vertex split. We derive a Gaussian information lower bound and an exact finite-support characteristic information that coincides with the bounded nonparametric KL-inf limit, and state sufficient conditions for first-order attainment of the Gaussian constant. Within the local-CS architecture, regularly varying boundaries imply a general allocation law with the 2/3 exponent for root-n variance-adaptive widths. Simulations and a PromptEval replay illustrate the resulting guarantee-efficiency frontier.
Read the full abstract ▾
Trading and Market Microstructure
A Generalized Langevin Model of Latent Liquidity and Concave Price Impact
Andrey Itkin
We model
market impact as the response to submitted order flow net of counterflow from latent traders, activated when price displacements from the level that would prevail without the order exceed individual thresholds. Order flow depletes this pool, and a generalized Langevin equation governs its recovery over several time scales. Its
memory kernels are finite sums of exponentials, so its Markovian lift is exact rather than an approximation. For an undepleted pool, aggregation under explicit assumptions on individual trading responses yields an intermediate square-root regime between linear small- and large-order limits, without imposing a square-root impact law. Scaling thresholds and responses with price noise makes impact in this regime proportional to
volatility, and thresholds that grow with the execution horizon make it independent of duration. With constant displayed depth, expected round-trip costs are nonnegative under the log-price convention, independently of the
memory. Numerical experiments show that depletion narrows the square-root range and that
memory spectra producing similar single-order impacts can respond differently after substantial prior trading. Calibration to
market data is left to a companion paper.
Read the full abstract ▾
Robotics
ExceptionDrive: A Planning-Oriented Counterfactual Corner-Case Benchmark for Autonomous Driving
Ziyi Luo, Zhe Sun, Yehao Lu et al. (+5 more)
Average performance on routine driving
benchmarks does not establish planner reliability under rare, safety-critical hazards. We proposed ExceptionDrive, a counterfactual planning
benchmark that uses VLM-assisted screening, localized multi-view editing, and quality
auditing to insert hazards into real nuScenes scenes while preserving their context. Its 21 tasks span six safety families and define hazard or conflict regions, local safety constraints, and acceptable responses. Because hazard insertion can invalidate the recorded human trajectory, our reference-free protocol evaluates edited predictions using Unsafe Rate (UR), Hazard Clearance Compliance (HCC), Hazard Proximity Response (HPR), and Counterfactual Trajectory Shift (CTS), which measure core-region intrusion, clearance compliance, clearance relative to a prescribed margin, and counterfactual trajectory change. Seven representative planners frequently intrude into hazard regions or provide insufficient clearance. We also develop a Reminder
Agent that, without sample-specific task labels, converts visual evidence and the shared taxonomy into structured records of hazard presence, type, and a recommended high-level strategy. The
agent neither predicts trajectories nor controls the vehicle; its records guide a VLM-based decision
agent. In zero-shot experiments, the reminders improve strategy accuracy and reduce under-warning.
Read the full abstract ▾
All authors Ziyi Luo, Zhe Sun, Yehao Lu, Lei Zhou, Lisheng Wu, Xuewei Li, Zequn Qin, Xi Li
Computer Vision
Learning from synthetic photorealistic raindrop for single image raindrop removal
Zhixiang Hao, Shaodi You, Yu Li et al. (+2 more)
Raindrops adhered to camera lens or windshield are inevitable in rainy scenes and can become an issue for many computer
vision systems such as autonomous driving. Because raindrop appearance is affected by too many parameters, therefore it is unlikely to find an effective model based solution. Learning based methods are also problematic, because traditional learning method cannot properly model the complex appearance. Whereas deep learning method lacks sufficiently large and realistic training data. To solve it, in our work, we propose the first photo-realistic
dataset of synthetic adherent raindrops for training. The rendering is physics based with consideration of the water dynamic,
geometric and photometry. The
dataset contains various types of rainy scenes and particularly the rainy driving scenes. Based on the modeling of raindrop imagery, we introduce a detection network which has the awareness of the raindrop refraction as well as its blurring. Based on that, we propose the removal network that can well recover the image structure. Rigorous experiments demonstrate the state-of-the-art performance of our proposed framework.
Read the full abstract ▾
All authors Zhixiang Hao, Shaodi You, Yu Li, Kunming Li, Feng Lu
Fluid Dynamics
A conservative bulk-surface closure for soluble-surfactant falling films: an analytical and numerical mass-balance audit
Sanghasri Mukhopadhyay and Asim Mukhopadhyay
Adsorption and desorption redistribute soluble surfactant between a
liquid film and its free surface, but cannot change the total inventory in a closed periodic domain. We audit this requirement in nonlinear weighted-residual models for soluble-surfactant-laden falling films. Although a previous correction restores conservation of the nonlinear surface Marangoni flux, a separate inconsistency remains in the depth-integrated bulk
diffusion. We derive the resulting mass imbalance and construct explicit positive periodic concentration fields that exhibit it. The missing contribution arises from the streamwise concentration gradient appearing in the normal diffusive flux at a sloping interface. Including this contribution restores local conservation of the coupled bulk-surface transport system while leaving the hydrodynamic equations, the leading-order adsorption-desorption closure, and the linear stability properties of the uniform
equilibrium unchanged. In an independent no-slip calculation using the parameters of an earlier travelling-wave study, the reference model loses $0.0228832\%$ of its initial surfactant mass by $t=4500$, whereas the corrected run has an absolute relative mass change of $1.2\times10^{-15}$. The main hump and capillary ripples remain closely comparable at $t\approx4500$. This
benchmark therefore shows that agreement with travelling-wave calculations does not, by itself, guarantee that a reduced model respects the conservation laws inherited from the underlying transport problem. Although the drift is small in this case, the corrected equations conserve the reduced total inventory for every smooth periodic solution. We also distinguish this exact conservation property from higher-order compatibility with the full curved-interface boundary conditions.
Read the full abstract ▾
Machine Learning
Learning Beyond What You Sample: Off-Policy-Aware Cross-Model Trajectory Exchange for RLVR
Doohyuk Jang, Yoonsik Park, Gyouk Chu et al. (+2 more)
Reinforcement Learning with Verifiable Rewards (RLVR) methods such as GRPO rely on successful self-generated trajectories, but finite rollout budgets can produce all-fail groups with no reward-based policy-gradient signal. While additional rollouts improve the chance of success at higher cost, successful trajectories missing from one model's rollouts may already have been discovered by another. Indeed, we observe that heterogeneous models often succeed on complementary
prompts, creating opportunities for mutual learning without a designated stronger teacher. To exploit this complementarity, we propose GRAFT (Gated Replacement of Answer-Failed groups with peer Trajectories), an off-policy-aware framework that replaces all-fail groups with informative peer groups. GRAFT transfers both successful and unsuccessful peer responses with peer-computed advantages, while controlling cross-model mismatch through sequence-level compatibility weighting and
token-level importance ratio clipping. Across three heterogeneous model pairs and five mathematical
reasoning benchmarks, GRAFT consistently improves both models over GRPO with the same per-model rollout budget, gaining 2.1 points on average and up to 4.5 points in model-level average performance. Stored peer trajectories preserve most of the gains, improving over GRPO by 1.8 points on average without simultaneous co-training.
Read the full abstract ▾
All authors Doohyuk Jang, Yoonsik Park, Gyouk Chu, Sihwan Park, Eunho Yang
Group Theory
Absolute moved spaces and noncrossing partition posets in arbitrary Coxeter groups
Thomas Gobet
The interval $[1,c]_T$ between the identity element and a Coxeter element $c$ in the absolute order on a Coxeter group $W$ is a generalization of the poset of noncrossing partitions arising when $W$ is the symmetric group. When $W$ is finite, this poset is always a
lattice, and it is natural to associate to every element $w\in [1,c]_T$ its moved space $\mathsf{Mov}(w)=\mathrm{Im}(w - \mathrm{Id}_V)$ in the
geometric representation $V$ of $W$. It has dimension equal to the reflection length $\ell_T(w)$ of $w$, and gives a realization of $[1,c]_T$ inside the
lattice of subspaces of $V$. It is an important tool in the study of $[1,c]_T$. When $W$ is infinite, the moved space of an element $w\in [1,c]_T$ no longer has dimension $\ell_T(w)$ in general, and distinct elements may have the same moved space. We propose a replacement for the moved space of an element $w\in [1,c]_T$ in an arbitrary Coxeter group, that we call absolute moved space of $w$. This subspace $\mathsf{AM}(w)$ of $V$ always contains $\mathsf{Mov}(w)$ and has dimension equal to $\ell_T(w)$, and distinct elements have distinct absolute moved spaces. This allows us to derive several properties of noncrossing partition posets that hold in full generality, and to show that the natural map from $[1,c]_T$ to reflection subgroups of $W$, which to $w\in [1,c]_T$ associates the subgroup $P(w)$ generated by reflections lying below $w$ in the absolute order, is always injective. Among others, we also derive a new proof of the
lattice property of $[1,c]_T$ when $W$ has rank three, and exhibit infinitely many new examples of infinite Coxeter groups of rank four and choices of Coxeter elements for which $[1,c]_T$ fails to be a
lattice.
Read the full abstract ▾
Statistics Applications
Feeling Left Behind? Territorial Disadvantage, Well-Being and Cohesion across European Regions
Stefano Maria Iacus, Giuseppe Porro, Haodong Qi et al. (+1 more)
Left-behind places are usually identified using economic, demographic and accessibility indicators, but these may not align with well-being and social cohesion. We combine ten years of territorial indicators with subjective measures derived from georeferenced social-media data for over 1,200 NUTS-3 regions in 28 European countries (2013-2023). Regional profiles and within-between panel models reveal uneven relationships: disadvantaged regions can display contrasting levels of well-being and cohesion, national context alters cross-country associations, and between-region differences diverge from within-region change. The findings show that aggregate measures conceal distinct forms and trajectories of territorial disadvantage.
Read the full abstract ▾
All authors Stefano Maria Iacus, Giuseppe Porro, Haodong Qi, Devika Jain
Machine Learning
Strict-Saddle Landscapes and Multi-Rank Geometry in Low-Tubal-Rank Tensor Sensing
Eugene Agyei-Kodie, Longxiu Huang, Shuang Li et al. (+1 more)
We study the optimization landscape of low-tubal-rank tensor sensing through a balanced factorization. Under a tubal restricted isometry condition, we establish a quantitative strict-saddle landscape with no spurious local minima for arbitrary
Fourier multi-rank profiles. We further show that the local geometry depends on the
Fourier-slice ranks rather than the tubal rank alone. Uniform ranks yield quadratic growth transverse to the solution orbit, whereas nonuniform ranks produce quartically flat directions through hidden frequency-wise overparameterization, even when the factor width equals the exact tubal rank. Numerical experiments illustrate the global optimization behavior and the contrasting local geometries.
Read the full abstract ▾
All authors Eugene Agyei-Kodie, Longxiu Huang, Shuang Li, Xiao Liang
Software Engineering
AgentBug-Smith: Automatically Reproducing Real-World Harness Bugs in Agentic Systems
Yiming Cheng, Alfin Wijaya Rahardja, Mengshi Zhang et al. (+3 more)
Agent harness bugs exhibit unique characteristics and remain challenging for state-of-the-art software
agents to repair. Progress in this area is further hindered by existing
benchmarks, which contain only a small and fixed number of executable harness bugs while requiring hundreds of human hours to construct. This work presents AgentBug-Smith, an automated harness bug reproduction approach that continuously discovers and reproduces real-world harness bugs from open-source agentic systems. Across different backbone LLMs, AgentBug-Smith consistently outperforms existing bug reproduction techniques designed for general software, achieving 10.67% - 27.56% higher success rates of reproducing harness bugs. By applying AgentBug-Smith to open-source agentic systems in the wild, we construct Live-Harness-Bench, a live and extensible
benchmark that currently contains 200 reproducible harness bugs. We further demonstrate the utility of Live-Harness-Bench through two downstream applications. First, we use Live-Harness-Bench as the evaluation
benchmark to systematically evaluate state-of-the-art software
agents, revealing their limited capabilities in repairing real-world harness bugs. Second, we use Live-Harness-Bench as a knowledge base of real-world harness bug fixes, from which reusable repair skills can be distilled to improve existing software
agents, increasing their harness-bug repair rates by 6.32%. Together, AgentBug-Smith and Live-Harness-Bench establish a scalable foundation for continuously evaluating and improving software
agents on harness bug repair, turning real-world
agent failures into executable evaluation instances and reusable knowledge for harness improvement, thus contributing to the ultimate goal of recursively self-improving
agents.
Read the full abstract ▾
All authors Yiming Cheng, Alfin Wijaya Rahardja, Mengshi Zhang, Zihao Chen, Zhenpeng Chen, Yiling Lou
Computation and Language
It's Not What the Image Shows: Irrelevant Context Destabilises VLM Judges Without Informing Them
Nagham Omar, Mahmoud Jabarin, Kinan Ibraheem et al. (+1 more)
Vision-
language models (VLMs) are increasingly used in place of human annotators, making it important that substitutability tests reflect the model rather than incidental evaluation conditions. We introduce MIST, the Misleading-Image Stress Test: 200 English sentences, each built around a phrase readable either figuratively or literally and shown with an aligned image depicting its reading, a misleading image depicting the opposite, or no image at all. The guidelines require the label to be decided from the sentence alone, so no image should change any answer. We expected each image to pull a judge's labels toward the sense it depicts, and neither kind did. Across thirteen VLM judges, an aligned image changed 20.5% of labels and a misleading one 19.4%, close for every judge and both above the 11.6% produced by deleting the ignore-the-image instruction with the image left in place. Yet only 37% of the labels that differ between the two images moved toward the sense shown, and agreement with our human annotators is unchanged whether the image is absent, aligned or misleading. The effect is smaller in the seven judges that pass the alt-test than in the six that never do, but present in all of them: what moves a judge is that an image is there, not which of the two it is, so a substitutability verdict describes a configuration as much as a model.
Read the full abstract ▾
All authors Nagham Omar, Mahmoud Jabarin, Kinan Ibraheem, Lotem Peled-Cohen
Algebraic Geometry
Isotropic localizations of mixed initial motives
Charles De Clercq and Fabio Tanania
In this paper, we define the triangulated category of initial motives, which contains, for example, all motives of projective homogeneous varieties for semisimple algebraic groups of inner type. By analyzing its Chow weight structure, we prove that the family of isotropic localizations indexed by finitely generated extensions of the base field is conservative and Pic-injective on initial motives. This result provides an extension, in the mixed setting, of the work of Quéguiner-Mathieu and the first author on higher Tate traces of Chow motives.
Read the full abstract ▾
Computation and Language
One Threshold Does Not Fit All Languages: Language-Conditional Deferral for Reliable and Efficient Low-Resource Text Classification
Bhanu Prakash Vangala and Vangala Navya
In the Global South, the lower-income countries of Africa, Asia, and Latin America where most of the world's
languages are spoken, a deployed text classifier usually runs on ordinary CPUs, serves many
languages with a single model, has few labeled examples in any of them, and relies on people to catch its mistakes. Such a system is only useful if it can promise how often it will be wrong: at most a fixed fraction of the labels it assigns on its own may be incorrect, and everything else must go to a person. Split conformal prediction delivers this promise through a single confidence threshold, normally estimated on validation data pooled across
languages. We ask whether the promise reaches every
language, and it does not. On MasakhaNEWS (16 African
languages) and AfriSenti (12
languages plus two never seen in training), a pooled threshold meets the 90% target on average but covers Somali at 77.5%, Tigrinya at 83.7%, and the two unseen
languages at 77.5% and 81.2%. Estimating one threshold per
language brings every
language to between 89.1% and 91.0% without retraining, and it shows how unequal the cost of the promise is: keeping it means sending 43% of Somali news and over 80% of Amharic and Xitsonga tweets to a person, against under 8% of Nigerian Pidgin news. One or two hundred labels per
language are enough and the models train in minutes on one CPU core, so the fix is affordable: calibrate, report, and budget human review one
language at a time.
Read the full abstract ▾
Analysis of PDEs
Finite-Time Vanishing of Vacuum and Large-Time Behavior of the Multi-Dimensional Degenerate Compressible Navier-Stokes Equations with Large Spherically Symmetric Initial Data
Qinghao Lei and Zhilei Liang
In this paper, we study the two- and three-dimensional barotropic compressible Navier-Stokes equations with degenerate density-dependent viscosity coefficients in the whole space or in a ball for arbitrarily large spherically symmetric initial data. For initial data allowing
vacuum, we establish the global existence of weak solutions and derive uniform-in-time a priori estimates. As a consequence, we prove that the
vacuum state of weak solutions will vanish in finite time. The key ingredients include a treatment of the pressure terms based on separate estimates in the low- and high-density regions, the Bresch-Desjardins
entropy estimates, weighted radial estimates, and a coupled control of the velocity and effective velocity. In the three-dimensional case, this approach also allows us to treat the endpoint case of the adiabatic exponent. For sufficiently regular initial data with strictly positive density, we establish the global existence and large-time behavior of classical solutions.
Read the full abstract ▾
Instrumentation and Methods for Astrophysics
An End-to-End Numerical Framework for Tidal Disruption Events with AthenaK
Hong-Xuan Jiang, Mengqi Yang, David A. Velasco-Romero et al. (+4 more)
In a
tidal disruption event (TDE), a black hole (BH) tears apart a
star, whose bound debris returns over many orbits to form an
accretion disk. We present a framework built on the GPU-accelerated finite-volume code AthenaK that follows these stages in a single simulation with self-
gravity throughout. We extend AthenaK with a moving frame, restart remapping between domains, an adaptive-mesh multigrid Poisson solver, a tabulated hydrogen–helium equation of state including recombination, a dual-energy update for the cold supersonic debris stream, a moving BH potential with excision, and localized adaptive time stepping (LAT), which speeds up the production run more than threefold. Each component is validated separately and in combination: the
gravity solver maintains an isolated Lane–Emden sphere to a density error of $2.0\times10^{-3}$ over 16.3 dynamical times, and dual-energy recovery reduces the pressure error of a Mach $7.75\times10^7$
entropy wave from 15.7 to $4.0\times10^{-12}$. We demonstrate the framework with a Newtonian $\beta=1$ disruption of a $1\,M_\odot$, $1\,R_\odot$
star by a $10^3\,M_\odot$ BH. The debris has the expected energy spread, insensitive to the self-
gravity update interval, splits evenly into bound and unbound material, and yields a fallback rate approaching $t^{-5/3}$ at late times. At the pericenter nozzle, the thermal energy gained in the high-resolution run matches the vertical kinetic energy lost to within 6%, whereas the fiducial run, with 4–8 times coarser
cells, overheats the thinnest early stream through numerical dissipation. Their heating agrees to 10% once the returning stream thickens. Multifrequency LTE post-processing produces synthetic images and luminosities.
Read the full abstract ▾
All authors Hong-Xuan Jiang, Mengqi Yang, David A. Velasco-Romero, Fangyuan Yu, Jing-Ze Xia, Xinyu Li, Yosuke Mizuno
Machine Learning
Storage Is Not Strategy: State-Conditioned Support Control for LLM Unlearning
Tianhao Qian, Ziming Hong, Chongyang Gao et al. (+2 more)
Many localized large
language model (LLM) unlearning methods select a small parameter subset from a localization signal and keep it fixed during optimization. The parameters most associated with a target, however, need not be the best ones to update, and candidate interventions can change value as optimization proceeds. In a controlled experiment, a storage-localization score reaches an area under the receiver operating characteristic curve (AUROC) of 0.981, yet storage identity agrees with the better intervention on only 17/36 targets, while low-rank adaptation (LoRA) wins 35/36. We introduce Intervention Score, which ranks editable groups by the predicted effect of the actual unlearning update while accounting for collateral damage, and use it to form the static intervention-value baseline (Static-IV). We then introduce selective dynamic intervention re-ranking (DIR-R), which revisits that subset only when a calibrated probe justifies the comparison. On the Natural-TOFU
dataset, our method has positive descriptive margins in 19/20 comparisons between methods and objectives, although several are near zero. On the LACUNA localization-precision
benchmark, our mean terminal utility is higher in all six negative preference optimization (NPO) and SimNPO comparisons: NPO margins range from +0.431 to +0.848, and SimNPO margins range from +0.503 to +0.571. The gradient-difference (GradDiff) objective reveals substantial field dependence. Relative to Static-IV, the primary four-field GradDiff evaluation has six wins, six ties, and no losses, with mean and median paired gains of +0.165 and +0.0025. The evidence supports separating localization, initial intervention selection, and checkpoint-dependent support revision.
Read the full abstract ▾
All authors Tianhao Qian, Ziming Hong, Chongyang Gao, Kezhen Chen, Lixu Wang
Artificial Intelligence
Active Budget Can Kill Sensitivity: Diagnosing and Repairing TopK Sparse Autoencoder Reliability
Zhenting Huang, Bo Jiang, Junnan Liu et al. (+2 more)
Sparse autoencoders (SAEs) are increasingly scaled to wider dictionaries to recover fine-grained structure from large
language model activations. However, a feature is useful for interpretation only if it remains a stable unit of analysis when the same meaning is expressed in different surface forms. We study this reliability question for TopK SAEs via feature sensitivity. Experiments demonstrate that scaling selectively reduces the sensitivity of rare features, while common features remain comparatively stable. A controlled width\(\times k\) factorial experiment identifies the active budget k as the root cause: the degradation arises from the selection boundary rather than dictionary width alone. We attribute this failure to the geometry of TopK selection. The active margin, the distance to the cutoff, predicts feature loss without thresholds. Guided by this margin diagnosis, we introduce pairwise rank stabilization. Our method targets ordering failures at the cutoff and improves rare-feature sensitivity by \(8.83\) percentage points, while keeping reconstruction and alive-feature coverage near the baseline. Overall, our results suggest that wide TopK SAEs should be evaluated not only by reconstruction, sparsity, and feature count, but also by feature reliability under semantic variation and boundary geometry for stable interpretability.
Read the full abstract ▾
All authors Zhenting Huang, Bo Jiang, Junnan Liu, Zhixing Tan, Qianren Mao
Plasma Physics
Effects of Alfvénicity on the Compressible Energy Cascade in Magnetohydrodynamic Solar Wind Turbulence: Evidence from Parker Solar Probe, Solar Orbiter, and WIND observations
Nahuel Andrés, Carlos A. Gonzalez and Norberto Romanelli
The compressible MHD energy cascade in the
solar wind is shaped by Alfvénicity,
plasma fluctuations, and
proton temperature, yet how these factors regulate the relative weight of compressible and incompressible energy transfer across heliocentric distances remains poorly understood. We apply compressible and incompressible exact relations to a large sample of 2-hour intervals from WIND (1~au), Parker
Solar Probe ($<0.25$~au), and
Solar Orbiter (0.3–1~au), classifying intervals as Alfvénic ($|\sigma_c|>0.75$) or non-Alfvénic ($|\sigma_c|<0.25$) and by wind speed, and examine the Yaglom-like and purely compressible components of the compressible cascade. Velocity fluctuation amplitude is the strongest predictor of the compressible cascade rate $\langle|\varepsilon_c|\rangle$, followed by density and
magnetic field fluctuations. $\langle|\varepsilon_c|\rangle$ and $\langle|\varepsilon_i|\rangle$ are tightly correlated in all regimes, with their ratio organized by Alfvénicity and wind speed: non-Alfvénic and Alfvénic fast winds cluster along $\langle|\varepsilon_c|\rangle=\langle|\varepsilon_i|\rangle$, while Alfvénic slow wind shows a systematic excess consistent with internal energy contributions. The largest $\langle|\varepsilon_c|\rangle$ systematically occur in hotter intervals, with a two-decade increase from cool to hot events in Alfvénic slow and non-Alfvénic wind, a trend carried mainly by the purely compressible term, while the dominant Yaglom-like term controls the overall cascade magnitude. Alfvénicity and wind speed thus govern the efficiency of turbulent energy transfer across heliocentric distances, and the tight coupling between compressible and incompressible rates points to the need for compressible, multi-fluid descriptions of the
solar wind dissipation budget.
Read the full abstract ▾
Computer Vision
HandAnthro: Automated Hand Anthropometry from a Single Image
Fan Zhou, Shuairan Chen, Mengying Zhang et al. (+6 more)
Hand anthropometry supports protective-glove design, but existing measurement methods often require trained operators, specialized hardware, or manual landmarking. We present HandAnthro, which estimates 44 projected hand dimensions from a smartphone photograph of a palm-up hand on US letter-size paper. The pipeline reconstructs wrist-occluded paper boundaries for rectification, whitens non-hand pixels, and refines 41 anthropometry-specific landmarks from a fine-tuned You Only Look Once (YOLO) pose model using image-specific geometry and contours. Controlled evaluation comprised 720 captures from 45 held-out participants, each contributing 16 images across two smartphones, two backgrounds, two angles, and two nominal illumination settings. HandAnthro produced complete outputs for 704 captures (97.8%); among these, mean absolute error (MAE) was 3.80 mm per dimension against two trained operators' caliper measurements. Regional MAEs were 2.48 mm for non-thumb fingers, 6.04 mm for thumbs, and 6.17 mm for palm and wrist. In a researcher-assisted mobile-app pilot, automated batch processing returned all 44 dimensions for 260 of 268 retained, researcher-screened firefighter images (97.0%). A descriptive, unpaired comparison with an independent national firefighter reference yielded a mean absolute difference of 2.40 mm across 28 sex-by-dimension group-mean contrasts. These results characterize controlled measurement performance and researcher-assisted field feasibility for future distributed hand-anthropometry studies.
Read the full abstract ▾
All authors Fan Zhou, Shuairan Chen, Mengying Zhang, Yulin Wu, Sadegh Jafari, Sixing Yu, Rui Li, Ali Jannesari, Guowen Song
Applied Physics
Combined photoluminescence and electrical characterization of valley photovoltaic devices explained with an equivalent circuit
Daixi Xia, Abhinav S. Sharma, Andreas Pusch et al. (+9 more)
We present an equivalent-
circuit analysis that simultaneously considers the photoluminescence and electrical responses of valley photovoltaic devices. Valley photovoltaics (VPV) are a novel concept for hot-carrier
solar cells, where intervalley scattering of electrons in the absorbing material helps maintain the conduction-band electron populations at metastable satellite valleys, which are at energies higher than the conduction-band minimum. We measure intensity- and voltage-dependent photoluminescence and electrical signals of a VPV device, which has previously shown S-shaped current-voltage characteristics and thus low efficiency. We present an equivalent
circuit that captures the S-shaped current-voltage curves and simultaneously describes our photoluminescence and electric measurements. The equivalent-
circuit model indicates that a reverse diode causes inefficient extraction of electrons and thus the S shape. We use a Poisson/drift-
diffusion model to demonstrate that the reverse diode can be either explained by the valley-scattering process itself or a heterojunction barrier. The reverse diode must be eliminated before VPV devices can achieve high efficiency.
Read the full abstract ▾
All authors Daixi Xia, Abhinav S. Sharma, Andreas Pusch, Murad J. Y. Tayebjee, Michael P. Nielsen, Fiacre E. Rougieux, Kyle R. Dorman, Tetsuya D. Mishima, Michael B. Santos, Ian R. Sellers, Nicholas J. Ekins-Daukes, Jacob J. Krich
Computation and Language
AnthroDial: Benchmarking LLM Anthropomorphism in Autonomous Social Interaction
Wentao Liu, Xi Chen, Siyu Song et al. (+13 more)
Large
language models (LLMs) are increasingly deployed as social
agents, yet credible human-like interaction requires more than fluent responses or persona consistency.
Agents must autonomously decide whether, when, and how to communicate while adapting to evolving contexts, goals, and relationships. Existing research, however, lacks a unified approach to enabling, evaluating, and improving such capabilities in continuous, open-ended interaction. We introduce AnthroDial, a unified framework for developing anthropomorphic social
agents from three complementary aspects: MindFlow, a lightweight interaction harness that enables autonomous, asynchronous, and adaptive communication through a dynamic Mind Buffer; CAPS-Eval, a theory-grounded framework for evaluating cognitive, affective, and behavioral dimensions of anthropomorphic interaction; and a scalable training paradigm that combines SEEDS for environment expansion with DiAPO for adaptive capability optimization. We further construct evaluation
datasets covering everyday communication, game interaction, and long-horizon character interaction. Extensive experiments across diverse models and scenarios demonstrate improved interaction autonomy and naturalness, validate the reliability, discriminativeness, and agreement with human rankings of CAPS-Eval, and confirm the effectiveness of our training paradigm. Together, these components provide a unified framework for developing credible human-like social
agents in open-ended interaction.
Read the full abstract ▾
All authors Wentao Liu, Xi Chen, Siyu Song, Biao Yuan, Yu Zhang, Zhou Zhuotong, Jingying Zhou, Guohao Feng, Shasha Hu, Tianfu Wang, Shangshang Yang, Haoyang Liu, Youjia Li, Xiaokun Wang, Min Ji, Ji Wang
Machine Learning
Delta-Matching: Closing the Final Gap of Native 8-bit Training for LLMs
Haozhan Tang, Hao Kang, Han Cai et al. (+2 more)
Reliable FP8
attention remains a barrier to fully native 8-bit large
language model training. We derive how forward-backward inconsistencies produce stale delta and empirically show how it distorts training dynamics. Our stale-delta hybrid runs show a modest loss gap at 569M parameters but substantial loss increases and downstream degradation at 1.67B and 5.29B. QK normalization, NoPE (no positional encoding), and lower-learning-rate context extension mitigate or delay degradation without eliminating it. This pattern suggests accumulated optimization error that smaller models and short runs can conceal. We propose Delta-Matching, proving that it restores the softmax gradient's zero-row-sum invariant under the stated numerical assumptions. It enables native block-scaled FP8 in every forward and backward
attention-core matmul without architectural changes, smaller global batches, or auxiliary forward outputs. Across tested architectures, scales, and training stages, Delta-Matching matches BF16/FP32 mixed-precision training loss and overall downstream performance. We will release our implementation, trained models, and data recipes (https://jet-ai-projects.github.io/delta-matching).
Read the full abstract ▾
All authors Haozhan Tang, Hao Kang, Han Cai, Song Han, Chenyan Xiong
Computer Vision
FlowMap-OPD: Rollout–Kernel Separation for On-Policy Distillation of Few-Step Flow-Map Generators
Zhiqi Li and Bo Zhu
Few-step flow-map generators, including MeanFlow and consistency models, enable efficient sampling through long-range transport, yet their on-policy
distillation remains underexplored. We introduce FlowMap-OPD, an on-policy
distillation framework that separates student-state acquisition from teacher–student distribution comparison. A formulation based on state marginals establishes this separation, while flow–velocity consistency connects local supervision to the deployed long-range map. Within this framework, we develop flow-map, induced-velocity, and instantaneous-velocity distribution supervision, each paired with a separately specified native flow-map rollout. Cross-capacity ImageNet experiments across three teacher rewards identify instantaneous-velocity distribution supervision with independently tunable student consistency as the most effective choice. In text-to-image experiments, FlowMap-OPD demonstrates strong multi-specialist consolidation capabilities and surpasses multi-reward Flow-Map GRPO in task performance and convergence speed.
Read the full abstract ▾
Computer Vision
RelayVSR: Large-Small Model Collaboration for Efficient Real-World Video Super-Resolution
Xijun Wang, Xin Li, Zirui Lang et al. (+3 more)
Large generative models can recover realistic detail in real-world
video super-resolution (VSR), but processing an entire
video with them is computationally expensive. In this work, we present RelayVSR, a streaming VSR framework built on the Sparse Generative Relay mechanism. A large generative model generates reference latents for sparse keyframes, while a lightweight VSR network uses these references and low-resolution
video to super-resolve every frame. The lightweight VSR network, implemented as a Dual-
Memory Video Transformer, reuses keyframe information across frames and updates recent
video context, supporting first-keyframe conditioning and dual-endpoint conditioning with bounded lookahead. However, errors in shared keyframes can propagate and accumulate across output frames, making keyframe quality alone an insufficient optimization target. We address this collaboration gap with
Video-Aware Reference Optimization (VARO), which uses reinforcement learning to update the large generative model with two reward levels: a system-level reward evaluates
videos produced by the fixed lightweight VSR network, while a reference-level reward evaluates decoded keyframe quality. VARO improves final
video quality over direct joint training, and its dual-level rewards outperform a system-level reward alone. At 1080p on a single NVIDIA A100 80GB, dual-endpoint RelayVSR with a 15-frame keyframe interval reaches 29.29 FPS, 13.82 GB peak GPU
memory, and 0.327 s first-frame model latency, compared with 7.80 FPS, 24.447 GB, and 2.83 s for FlashVSR-Tiny. The code is available at https://github.com/kopperx/RelayVSR.
Read the full abstract ▾
All authors Xijun Wang, Xin Li, Zirui Lang, Suhang Yao, Haoran Li, Zhibo Chen
Software Engineering
Is manual software optimization a thing of the past?
Pavlin G. Poličar, Martin Špendl and Tomaž Hočevar
Scientific software is increasingly required to process larger
datasets while maintaining acceptable execution times. Software optimization traditionally requires substantial expertise in programming,
algorithms, and numerical methods. Recent advances in large
language models (LLMs) offer the possibility of automating much of this process. We investigate whether LLM-based
agents can autonomously achieve substantial performance improvements in scientific software, including mature implementations that have already been extensively optimized by human developers. We tasked an LLM-based
agent with optimizing software for three computational problems: t-SNE, single-sample gene set enrichment analysis (ssGSEA), and graphlet counting. Humans defined the scope, correctness criteria, and a verification mechanism, after which the
agent worked autonomously, in some cases for several hours. Code maintainers reviewed each resulting implementation and verified its correctness. The optimized implementations were faster in all tested configurations, by up to two orders of magnitude over the fastest existing tools. The improvements included low-level code optimizations, mathematical reformulations, and an entirely new
algorithm for graphlet counting. Software optimization can increasingly be delegated to autonomous
agents, with the human role shifting from implementing optimizations to deciding which software to optimize, defining objectives, providing verification mechanisms, and ensuring the correctness of the final software. For well-scoped, verifiable problems, we argue that manual software optimization may be a thing of the past.
Read the full abstract ▾