Peer Revenue Publish or profit

Papers

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

All of arXiv

Saturday 5 September

Computation and Language

SurveyAgent-HKA: A multi-agent framework for scientific survey generation with LLMs and human knowledge augmentation

Automatic scientific survey generation has become an important task in scientific document processing. The common approach of retrieving literature from a single source (e.g., arXiv) and generating surveys through a one-pass large language model (LLM) call often leads to limited reference coverage and, more importantly, fails to replicate the expert-driven revision process that is crucial for writing high-quality surveys. In this paper, we introduce SurveyAgent-HKA, a multi-agent framework that improves end-to-end scientific survey generation by incorporating knowledge derived from published surveys and peer-review comments. The framework decomposes survey generation into well-defined sub-tasks handled by LLM-powered agent. It first retrieves relevant papers from multiple sources and identifies key topics through clustering to construct an initial outline, which is then refined using outlines from related human-written surveys. Based on the refined outline, topic-focused papers are retrieved and re-ranked to select for drafting a well-grounded survey. Then, we identify common issues raised by experts in peer-review comments from published surveys to guide the revisions and finalize the survey. Experiments on two domains show that our approach outperforms mainstream baselines in citation quality, structural consistency, and content quality. Furthermore, our framework is efficient in both time and cost, making it a practical solution for broader AI-assisted scientific writing applications.
Read the full abstract

All authors Tong Bao, Mir Tafseer Nayeem, Yi Zhao, Davood Rafiei, Chengzhi Zhang

Computer Vision

Beyond Classification: Structured Supervision Aligns Visual Evidence with Medical Semantics

Vision Transformers (ViTs) have shown immense potential in medical image analysis. However, standard pre-training via global image classification suffers from spatial collapse, where models rely heavily on background shortcuts rather than localising critical foreground lesions. To overcome this limitation and align visual evidence with precise medical semantics, we systematically investigate alternative pre-training paradigms.Specifically, we evaluate three independent forms of structured supervision: topological priors via graph self-supervision, dense pixel-level constraints via segmentation, and cross-modal semantic grounding via image-text pairs. Notably, our empirical analysis reveals that while all three forms of structured supervision successfully alleviate the global pooling bottleneck and steer visual attention towards foreground regions, image-text alignment achieves the most superior performance. By embedding high-dimensional diagnostic logic, the cross-modal approach not only anchors attention on precise visual evidence but also enables profound abstract reasoning. Extensive experiments demonstrate that this semantically enriched pre-training fundamentally enhances the model's feature representation. Consequently, when fine-tuned for downstream clinical classification tasks, our models achieve superior accuracy and yield highly interpretable attention maps focused on true pathological features, vastly outperforming vanilla classification baselines.
Read the full abstract

All authors Hexiang Bai, Hanyang Xu, Xiaoxue Li, Xiaoliang Wu, Shangde Gao, Hongxia Xu, Ke Liu

Combinatorics

The Negami Polynomial and Broken-Circuit Stanley–Reisner Rings

We summarize the classical passage from the chromatic specialization of Negami's polynomial to the Hilbert series of a broken-circuit Stanley–Reisner ring, and record formulas for several graph families with their connectivity hypotheses made explicit. Our main construction in this revision is the associated graded algebra of the squarefree edge algebra under the graphic-matroid rank filtration, for an arbitrary finite loopless graph. Its basis is indexed by all edge subsets, and its multiplication is governed by disjointness and rank additivity. Its bigraded Hilbert polynomial recovers the full three-variable Negami polynomial. We prove that its socle has a basis indexed by edge subsets of full matroid rank; for a connected graph, these are precisely the connected spanning subgraphs. A vertex-degree refinement counts connected regular spanning subgraphs. This finite-dimensional algebra is distinct from a broken-circuit Stanley–Reisner ring. We distinguish these elementary structural results from the unresolved question of whether abstract algebra isomorphism can distinguish graphs with the same Negami polynomial; no priority claim is made for the construction or its Hilbert identity.
Read the full abstract

Rings and Algebras

Integral Dependence over One-Sided Ideals, Part I: Foundations and Classical Applications

We introduce notions of integrality and full integrality over one-sided ideals, together with associated radical constructions, and develop a general framework for their study. As applications, we extend several classical results concerning Jacobson radicals to the one-sided setting. In particular, we prove that the Jacobson radical of a left ideal is integral over that ideal in left artinian rings and in algebraic algebras, obtaining a one-sided analogue of a theorem of Amitsur. We also show that Jacobson radicals of graded left ideals are graded and derive a corresponding extension of Amitsur's theorem on Jacobson radicals of polynomial rings.
Read the full abstract

Combinatorics

Algebraic characterizations of generating and affinely generating $\Gamma$-magic maps and $\Gamma$-distance magic labelings on regular graphs

For an abelian group $\Gamma$ of order $n$, a graph $G$ of order $n$ is $\Gamma$-distance magic if it admits a bijection $V(G)\to\Gamma$ whose open-neighborhood sums are constant, and group distance magic if this holds for every such $\Gamma$. More generally, for any finite abelian $\Gamma$, a $\Gamma$-magic map is a map $f\colon V(G)\to\Gamma$ with constant open-neighborhood sums; we call $f$ generating if its labels generate $\Gamma$, and affinely generating if its pairwise differences do. Let $G$ be regular, let $\Gamma\cong\mathbb{Z}/d_1\oplus\cdots\oplus\mathbb{Z}/d_r$ with $d_1\mid\cdots\mid d_r$, put $B_\Gamma=\bigoplus_{i<r}\mathbb{Z}/d_i$, and let $\overline{A}_m$ be the adjacency operator induced on $(\mathbb{Z}/m)^{V(G)}$ modulo constants. We prove that $G$ admits a generating $\Gamma$-magic map iff $B_\Gamma\hookrightarrow\ker\overline{A}_{d_r}$, and an affinely generating one iff $\Gamma\hookrightarrow\ker\overline{A}_{d_r}$; when $|V(G)|=|\Gamma|$, it is $\Gamma$-distance magic iff the latter embedding has vertex-separating image. If the reduced adjacency operator is nonsingular over $\mathbb{Q}$, these become subgroup conditions in the reduced adjacency Smith group. Cichacz and Froncek conjectured that every distance magic graph is group distance magic. Using the affine criterion we construct a $6$-regular distance magic graph of order $27$ admitting no affinely generating $(\mathbb{Z}/3)^3$-magic map, so the conjecture fails even for affine generation. We propose the generating group distance magic conjecture, and prove it for regular distance magic graphs whenever $\Gamma$ is $2$-generated, hence for cube-free order. Further applications concern Cayley graphs on elementary abelian $p$-groups, Hamming relation graphs, strongly regular graphs, and symmetric designs.
Read the full abstract

Machine Learning

Rethinking the Evaluation of Efficiency Methods for Multi-Agent Systems

Efficiency is increasingly important for Large Language Model (LLM)-based multi-agent systems (MAS), as larger models and more agents introduce substantial execution costs. Recent methods aim to make MAS cheaper by pruning agents, removing communication edges, or searching for compact structures. However, we argue that existing evaluations may overestimate their true ability to improve MAS efficiency. Reported gains are often measured under method-specific prompts and starting topologies, making them difficult to attribute to the proposed structural changes. Moreover, many reported successes appear in non-MAS-demanding settings, where a single agent or a randomly pruned system can already preserve strong performance. To study these issues, we introduce a controlled and MAS-demanding diagnostic benchmark for representative MAS efficiency methods. We evaluate methods under a shared backbone model, agent registry, and runtime, across controlled variations in topology, scale, depth, and tool use. Our analysis shows that many reported gains are setup-dependent and may arise from structural collapse, disabled tool pathways, or starting systems where random pruning already preserves accuracy, rather than robust improvements in MAS efficiency.
Read the full abstract

All authors Jiamu Zhang, Lingxi Zhang, Pengjun Lu, Qiyue Zhang, Yu-Neng Chuang, Zhengchen Li, Shuai Xu, Vipin Chaudhary, Hanjie Chen

Rings and Algebras

Finite-Sample Rigidity for Left-Translated Similarity Orbits and Off-Diagonal Quadratic Realizations

We study finite-sample rigidity for left translates of similarity orbits over commutative rings. Let \(R\) be a commutative ring with \(2\in R^\times\), and let \(C\in\operatorname{Mat}_2(R)\) be cyclic. Given matrices \(G_i=FK_i\), where \(F\in\operatorname{GL}_2(R)\) is common and the underlying points \(K_i\) lie in \(\operatorname{Orb}(C)\), we determine when a finite translated sample has exactly the same multiplier ambiguity as the complete translated orbit. The key structural result is that the orbit differences span \(\mathfrak{sl}_2(R)\), which yields a scalar description of the left stabilizer. We then give a determinantal saturation criterion, expressed directly in the matrices \(G_i\), under which finite-sample consistency reduces to trace equations and one determinant condition. Off-diagonal quadratic operators on pairs of free rank-two modules realize the translated-orbit model exactly. Over fields we obtain sharp sample thresholds under the saturation criterion, and over finite fields we derive exact saturation probabilities for regular semisimple and nonzero nilpotent orbits. Torsion restrictions of elliptic isogenies provide a natural arithmetic realization of the rank-two module framework.
Read the full abstract

Medical Physics

Distinct Radiobiological Responses to BNCT in SAS Oral Squamous Cell Carcinoma and MCF-7 Breast Cancer Cells

This work compared the radiobiological responses of SAS oral squamous cell carcinoma cells and MCF-7 breast cancer cells following accelerator-based boron neutron capture therapy (BNCT). Neutrons were generated by bombarding a lithium target with proton beams, followed by moderation to obtain sufficient thermal neutrons for BNCT irradiation. Boronophenylalanine (BPA) was used as the boron delivery agent. BNCT-induced biological responses were evaluated by gamma-H2AX immunofluorescence staining, cell-cycle analysis, apoptosis analysis, and clonogenic survival assays. BNCT induced marked gamma-H2AX foci formation in both cell lines, indicating DNA damage-associated responses after irradiation. The two cell lines further showed distinct post-irradiation outcomes. SAS cells exhibited stronger clonogenic suppression and prominent G2/M accumulation, whereas MCF-7 cells showed sustained G0/G1 accumulation and delayed apoptosis. These results suggest that BNCT sensitivity is determined by both boron accumulation and cell-line-specific biological characteristics. This work provides experimental evidence highlighting the importance of tumor-dependent cellular responses in understanding and optimizing BNCT efficacy.
Read the full abstract

All authors Yuxiang Zhao, Zhao Sun, Changming Wang, Jianghao Lai, Jie Zhou, Zhencen He, Zhimin Hu

Analysis of PDEs

Sharp dispersive estimates for wave equation on lattice graphs with high dimension

Schultz [Comm.~Pure~Appl.~Math., 1998] first established dispersive estimates for the fundamental solution of the discrete wave equation on lattice graphs $\Z^d$ for $d=2,3$, which can be seen as a discrete analogue of the classical dispersive estimate for the Euclidean wave equation. His result was extended by Bi, Cheng and Hua to $d=4$ and $5$. In this paper, we give complete answers to all the remaining dimensions, including $d=5$. Our estimate is sharp for every $d\geq 5$. Moreover, the proof remains valid for the upper bound of the case for $d=4$.
Read the full abstract

Category Theory

Cartesian closedness of the category of real-valued sets, II

Let $[0,1]_*$ be the unit interval $[0,1]$ equipped with a left-continuous t-norm $*$. We prove that $[0,1]_*\text{-}\mathbf{Set}$ is cartesian closed if and only if $*$ is the minimum t-norm. This extends the classification for continuous t-norms established in the first part of this work to the whole left-continuous setting.
Read the full abstract

Computation and Language

Solving versus Verifying: Catching Contradictions in Tax Reasoning Systems

Large language models now compute correct tax liabilities on over 90% of well-formed cases in statutory benchmarks, which makes them candidates for the tax-advisory and compliance systems that consume such an answer directly. Real legal inputs, however, are frequently defective: required facts are missing, or stated facts contradict one another. Accuracy on clean benchmarks says nothing about how a model behaves then, and a system that computes straight through a defective input returns a confident number with no sign that anything is wrong. This raises two questions: does a model asked to solve a case abstain when the input is defective, and when it does not, can the same model catch the defect when asked instead to verify the input? We study six recent models on SARA-derived tax cases under missing-fact and contradictory-fact perturbations. The strongest models abstain when a fact is missing but compute through injected contradictions, returning the clean-input answer 63-76% of the time with no signal of the conflict; asked instead to verify the same input, they flag most of those contradictions. We wire that verification call into a simple contradiction gate: one extra call that abstains when the model reports a conflict. Across all six models it recovers most of the missed contradiction abstention at a clean-accuracy cost of at most about 5 percentage points, with no training and no external tooling. High accuracy on well-formed inputs is therefore an incomplete measure of reliability, and the detection the solver misses is cheaply recoverable with a single self-check.
Read the full abstract

Robotics

GIF: Agentic Generation of Interactive and Functional Object Compositions for Robot Learning

Robot manipulation foundation models require scalable evaluation and data generation across diverse scenarios, with simulation providing an environment for both. Automated scene generation offers a promising path, yet prior work has largely emphasized coarse-grained scene layouts rather than fine-grained functional object compositions. Motivated by this gap, we present GIF, an agentic Generation framework for Interactive and Functional object compositions. In this framework, we recast this problem as disentangled reconstruction followed by relative pose recovery. CoGen produces instance-disentangled meshes with coarse initial poses leveraging complementary strengths of 2D and 3D generative models. GPRM refines the relative pose under joint geometric and physical guidance, and a VLM verifier selects the candidate that best matches the structured specification. We further construct a benchmark spanning eight representative contact-geometry classes and compare with state-of-the-art generators; GIF improves both asset quality and relation matching, while reducing collision rate to below 1%. Finally, we synthesize data for policy learning, revealing diversity scaling in both simulation and real-world deployment.
Read the full abstract

All authors Long Xu, Zhiqi Zhang, Mi Yan, Shengliang Deng, Chong Xia, Mingyu Dong, Jiayi Chen, Jiangran Lyu, Fei Gao, Zhizheng Zhang, He Wang

Algebraic Geometry

Open-closed duality in higher genus and winding

Let $X$ be a toric Fano surface. Let $\pi: \widehat{X} \rightarrow X$ be a toric blow up at a point. We use the Topological Vertex [AKMV] to prove a higher genus, higher winding, open-closed duality between the toric Calabi-Yau 3-folds $K_X, K_{\widehat{X}}$. For $g \geq 0, w \geq 1$, we show the equality $n_g(K_{\widehat{X}}, \pi^*\beta - wC) = (-1)^g N_{g, (w)}^{LMOV}(K_X/L, \beta)$, where $n_g(K_{\widehat{X}}, \pi^*\beta - wC)$ is the genus-$g$, Gopakumar-Vafa invariant of $K_{\widehat{X}}$ in curve class $\pi^*\beta-wC$, where $\beta \in H_2(X, \mathbb{Z})$ and $C$ is the exceptional curve, and $N_{g, (w)}^{LMOV}(K_X/L, \beta)$ is the genus-$g$, LMOV invariant of an outer Aganagic-Vafa brane $L \subset K_X$ in class $\beta$ and representation $(w)$, or the Young Tableau of a single row with $w$ boxes.
Read the full abstract

Computer Vision

AVSplat: Dense-View Feed-Forward 3D Gaussian Splatting with Assist-View Preconditioning

Pose-free feed-forward 3D Gaussian Splatting enables novel view synthesis from uncalibrated multi-view images. Although more views should improve performance, existing methods often degrade with dense-view inputs because global aggregation spreads attention over many tokens, and naive voxel fusion averages many Gaussians into overly smooth representations. We present AVSplat, a framework that turns additional views into reliable signals for both aggregation and representation. Before global attention, each view performs a single lightweight interaction with a small set of Assist Views chosen for relevance and diversity, and the cached features provide a focused scene context that stabilizes correspondence. For representation, we use adaptive temperature-aware voxel fusion that sharpens attribution under high occupancy, guided by occupancy and point confidence. Crucially, AVSplat restores positive view scaling where performance remains stable or improves as more input views are added, instead of degrading in the dense-view regime. Ablations show that Assist View Preconditioning is primarily responsible for preventing dense-view degradation, while Occupancy-guided Voxel Fusion contributes most of the single-point image-quality gains.
Read the full abstract

All authors Muyu Xu, Fangneng Zhan, Yu Wei, Hanspeter Pfister, Shijian Lu

Geophysics

Conservation of Normalized Energy Density in multi-layered media: a proof for the general multi-layer 2D SH problem

The Normalized Energy Density (NED) was introduced in Goto et al.(2011) as a quantity that, unlike the conventional energy, is conserved across material interfaces in a layered elastic medium. Its conservation was proved analytically only for a two-layer structure; for three or more layers it was supported by Monte Carlo simulation. I supply the missing proof for a general multi-layer (n-layer) structure in the 2D SH problem. The key step is to stop treating the angular frequency as a single variable and instead lift the layer phases to independent coordinates on the n-torus. The obstruction identified in the original paper—that the ratio of travel-time combinations is generically irrational, so that the periodicity argument used for two layers fails and disappears, and the proof reduces to Weyl's equidistribution theorem combined with n repetitions of the same elementary integral used in the two-layer case. I prove conservation in two independent settings that together cover all cases of practical interest: travel times that are linearly independent over Q, and commensurable travel times, including the equal-travel-time (Goupillaud) discretization. I also show that an arbitrary layered structure can be subdivided by reflectionless interfaces into one whose travel times take only d distinct, Q-independent values, where d is the dimension of the Q-span of the original travel times; this places the two proved cases at the endpoints d=1 and d=n and isolates the remaining open case 1<d<n in a canonical form.
Read the full abstract

Nuclear Theory

Effective kinetic theory description of the magnetic field-induced anisotropic gluon pressure during pre-equilibrium in heavy-ion collisions

We develop an effective kinetic description for the interaction of gluons and magnetic fields during the pre-equilibrium stage of relativistic heavy-ion collisions. For this purpose, we formulate the Boltzmann-Vlasov equation with the interaction term modeled by the effect of the magnetic field on the elements of an electrically charged and colored dipole originating from the quantum fluctuation of gluons into quark-antiquark pairs. We find the numerical solution of the collisionless Boltzmann-Vlasov equation with a time-dependent magnetic field profile to determine the time evolution of the directional pressures. The presence of the magnetic field tames the growth of the transverse to longitudinal pressure ratio compared with the case in the absence of the magnetic field; however, the system does not isotropize since collisions are not included. We study the cases of initial gluon distributions with longitudinal anisotropies, as well as the case of an initial isotropic gluon distribution. In both cases, we find that the magnetic field produces a non-monotonic early-time response for large values of the initial magnetic field strength.
Read the full abstract

All authors Khwahish Kushwah, Alejandro Ayala, Gabriel S. Denicol, Ana Mizher

Geophysics

Seismic P-wave attenuation estimation based on frequency-dependent AVO using Kramers-Kronig relations for gas reservoir prediction

Estimation of seismic attenuation (inverse quality factor) is important for gas reservoir prediction. Two key issues in seismic attenuation estimation are the development of a physically consistent reflection coefficient equation and stable estimation of seismic attenuation from seismic data. To address these issues, this study, within the framework of isotropic linear viscoelastic media, starts from the Kramers-Kronig relations and expresses the viscoelastic stiffness matrix as a function of seismic attenuation. Under the assumptions of weak attenuation and small elastic and attenuation contrasts across the interface, a frequency-dependent PP-wave reflection coefficient equation explicitly containing seismic attenuation terms is derived using scattering theory. The derived reflection coefficient equation not only satisfies the causality constraint, but also preserves a compact mathematical form. Based on this reflection coefficient equation, seismic attenuation is estimated within the framework of frequency-dependent AVO inversion. Synthetic seismic data tests show that the estimated P-wave attenuation attribute is sensitive to variations in reservoir gas saturation, with reservoirs of higher gas saturation exhibiting stronger P-wave attenuation anomalies. Application to field seismic data further demonstrates that the P-wave attenuation anomalies agree well with the gas saturation log and effectively identify high gas saturation reservoirs. This study provides a new approach for extracting P-wave attenuation information from seismic data and achieving high resolution prediction of gas reservoirs.
Read the full abstract

All authors Shengyi Wang, Xuehua Chen, Xingyu Luo, Kaixing Huang

General Relativity and Quantum Cosmology

Slow rotation of a localized Kalb-Ramond wormhole: Wald charges and fractional quadrupolar hair

We construct an action-consistent slow-rotation branch, through second order, for a traversable wormhole supported by a phantom scalar, a scalar-localized radial string sector, and a Kalb-Ramond two-form coupled nonminimally to both the Ricci scalar and Ricci tensor. For the regular benchmark branch, covariant asymptotic analysis shows a vanishing ADM mass, while the first-order axial mode carries finite Iyer-Wald angular momentum. The nonminimal contributions to the angular charge decay at infinity, so the covariant and metric definitions of angular momentum agree. At second order, no nonzero mass correction is resolved; at fixed asymptotic scalar charge, rotation increases the throat area and makes the throat oblate. The most distinctive result arises in the quadrupolar sector. A long-range polar Kalb-Ramond response couples to the localized background and generates a matter-supported metric tail that decays more slowly than the ordinary vacuum quadrupole. The same noninteger hierarchy appears in curvature and reduces the differentiability of the conformal completion at spatial infinity. Consequently, the Geroch-Hansen and asymptotically Cartesian mass-centered quadrupole constructions are not applicable in their standard smooth form to this branch, even though its global mass and angular-momentum charges remain well defined. This separates charge-level asymptotic flatness from smooth multipolar asymptotics and identifies fractional Kalb-Ramond hair as the leading rotational quadrupolar signature of the benchmark solution.
Read the full abstract

Cryptography and Security

Versioned Transitive Dependency-Closure Binding and Operation-Time Effect Governance for Agent Skills: ClosureBound

Agent Skills combine instructions with files, packages, tools, models, and services, so operational identity can exceed a signed directory. Recursive or lazy dependencies may change while root-level evidence remains valid, and different surfaces may reach the same durable effect. We present ClosureBound, a reference monitor that prevents authorization transfer across material changes to this heterogeneous closure. Its resolver commits typed graph nodes and topology. Each grant binds an exact closure root, effect ceiling, purpose/provenance, validity, and epochs. At durability, it re-resolves closure and state, normalizes the operation into an external-effect IR, and admits it only if a joint witness satisfies every bound. Supported equivalent paths share one ceiling. Assuming complete mediation and discovery, authenticated freshness, sound normalization, cryptographic binding, and authoritative linearization, we establish metadata non-authority, closure determinism, version non-inheritance, effect non-amplification, bound-value freshness, and path invariance. We do not establish program equivalence or remote-service honesty. A provider-free implementation matches 40 frozen lifecycle fixtures; 18 kernel contracts and six mutants cover binding and downgrade cases. Full-profile exploration reaches 84,608 states and 530,752 transitions without a declared invariant violation; six weakened profiles yield witnesses. A lexical audit of 549 public Skills (4,872 unique files) finds that 21 of 526 roots with bundled files name every non-manifest path verbatim, 67 contain links resolving outside their roots, and no root declares a frontmatter dependencies field. These observations motivate conservative closure discovery and define concrete targets for broader runtime, interoperability, efficacy, and production validation.
Read the full abstract

Machine Learning

A dictionary learning framework for graphs via filters and optimal transport

We propose a graph dictionary learning (GDL) framework where each graph is represented as a zero-mean Gaussian distribution derived from its filtered Laplacian. Each observed graph is approximated by a barycenter over learned atom graphs, computed under the filter graph distance (fGOT), a graph comparison metric sensitive to global structural properties. The reconstruction error between the observed graph and its barycenter is measured by the surrogate fGOT (sfGOT) distance, a tractable approximation of fGOT that handles graphs without known node correspondence, and is minimized end-to-end via backpropagation. We further provide a novel interpretation of sfGOT through the lens of the Hilbert-Schmidt Independence Criterion, showing that minimizing the sfGOT distance between two graphs is equivalent to maximizing statistical dependence between the spectral embedding of their nodes. Experiments on benchmark datasets demonstrate competitive performance over existing GDL methods on graph clustering and classification tasks.
Read the full abstract

Computer Vision

PAI-Actor: Cinematic Multi-Character Replacement in Dynamic Scenes

We present PAI-Actor, a cinematic multi-character animation framework for character replacement in dynamic movie scenes. Unlike conventional animation systems that mainly drive a single static image or a single subject, our goal is to replace and animate multiple characters within real video clips while preserving the original scene dynamics, camera motion, and background content. This setting is particularly challenging because the generated characters must remain consistent with the source performance in motion and interaction, while also matching the surrounding background in lighting, shadow, composition, and overall cinematic appearance. To address this, we formulate multi-character animation as a structure-guided human recovery problem and build a movie-driven training pipeline from high-quality film data. Furthermore, to support practical cinematic production, we introduce a bidirectional-to-autoregressive distillation framework: we first train a bidirectional diffusion transformer for high-quality short-clip generation at 1080P resolution, and then distill it into an autoregressive video-to-video model for efficient inference and longer video generation. Experiments show that PAI-Actor enables high-fidelity multi-character animation with strong scene consistency, cinematic visual quality, and efficient long-form generation.
Read the full abstract

All authors Bangxun Tang, Heyuan Gao, Yiren Song, Guian Fang, Zijian He, Jie Yang, Mike Zheng Shou

Combinatorics

Three Infinite Classes of APN Permutations on $Z_n$

For any permutation of a nontrivial finite abelian group, the differential uniformity is at least two; permutations attaining this bound are called almost perfect nonlinear (APN). We construct three infinite classes of APN permutations on the cyclic group $\mathbb{Z}_n$ using Singer cycles, binomials inducing projective permutations, and completed reciprocals combined with parity and quadratic characters. The respective domain orders are $q+1$ for prime powers $q>2$, $(3^d-1)/2$ for integers $d\ge2$, and $2p$ for primes $p>5$ with $p\equiv5\pmod6$. Each class contains an infinite subclass of composite orders outside the standard forms $r-1$, $r-2$, $r-3$, and $r-4$, where $r$ is a prime power. These forms arise in the Welch–Costas, Panario–Sakzad–Stevens–Wang, and Golomb constructions. To the best of our knowledge, these are the first infinite APN constructions on $\mathbb{Z}_n$ reported since 2011 that yield infinitely many composite orders outside these standard forms.
Read the full abstract

Computer Vision

STAR-Pro: Stage-Wise Token Adaptive Reduction with Progressive Refinement for Efficient Large Vision-Language Models

Large vision-language models (LVLMs) achieve strong multimodal understanding, but the hundreds to thousands of visual tokens they process impose substantial computational overhead, motivating training-free visual token pruning. In this work, we conduct two complementary analyses of visual token pruning. First, we measure the feature-space coverage of tokens retained before cross-modal fusion and find that aggressive pruning discards substantial visual information. Second, we track text-to-visual attention across decoder layers and find that the visual tokens considered important change substantially with depth, making one-shot pruning decisions unreliable. Together, these findings show that effective pruning should preserve broad visual coverage before fusion and progressively refine the retained tokens as cross-modal evidence evolves during fusion. We therefore propose STAR-Pro (STage-Wise Adaptive Token Reduction with Progressive Refinement), a training-free two-stage framework. Its Adaptive Stage applies pivoted QR to construct an over-budget feature-coverage candidate pool, while its Progressive Stage uses evolving text-to-visual attention at selected decoder layers to prune a nested survivor set under a target layer-average token budget. Extensive experiments across seven LVLMs spanning multiple architectures and 18 image and video benchmarks demonstrate the effectiveness of STAR-Pro under aggressive pruning. On LLaVA-Video-7B, STAR-Pro reduces visual tokens by 90.5%, retains 92.7% of baseline performance, and achieves a $2.24\times$ measured inference speedup. Code is available at https://github.com/EasonAI-5589/starpro.
Read the full abstract

All authors Yichen Guo, Tinghao Wang, Qizhe Zhang, Lingbei Meng, Yuan Zhang, Jiajun Cao, Hao Jiang, Chenwei Wu, Jixian Wu, Sixiang Chen, Tao Luo, Hongyang Cheng, Kai Tang, Chenxi Li, Renyuan Li, Xiande Huang, Wenya Wang, Shanghang Zhang

Signal Processing

Inverse Stochastic Filtering for Inverse Cognition in Counter-adversarial systems

Recent advances in cognitive and counter-adversarial systems with the ability to infer their adversary's beliefs have garnered significant research interest in inverse filtering from a Bayesian perspective. Such counter-adversarial systems have potential applications where an autonomous system interacts with its environment, learns relevant information about it, and then adapts itself to achieve its goals optimally. An important example is the inverse cognition problem in cognitive radar-target dynamics. A cognitive radar adapts its waveform, beam pattern, and other parameters in real-time to accurately localize its target. An intelligent target can employ these inverse filtering results to evade detection and even deceive the adversarial radar. In this setting, a cognitive adversary or 'attacker' tracks its target of interest using a stochastic framework such as a Kalman filter (KF). The target or 'defender' then employs another inverse stochastic filter to infer the forward filter's estimates of the defender's state computed by the attacker. For instance, recently, inverse KF (I-KF) has been developed to predict the attacker's future steps by estimating its (forward) KF's estimates. Existing works on inverse stochastic filtering and inverse cognition consider either linear Gaussian or finite state-space models, which limits their applicability in many practical engineering problems. Hence, in this dissertation, we study inverse filtering in the general non-linear non-Gaussian system framework and develop inverse non-linear filters using different sub-optimal approaches.
Read the full abstract

Probability

A Gamma envelope and sharp moment inequalities for Gaussian quadratic forms

We study extremal absolute central moments of Gaussian quadratic forms under a fixed Frobenius norm. For a nonzero real symmetric matrix $M$ and $G\sim N(0,I_n)$, we construct an explicit centered difference of Gamma variables with the same mean, variance, and third centered moment as $G^{\mathsf T}MG-\operatorname{tr}M$. After normalization, replacing the quadratic form by this Gamma difference does not decrease $\mathbb{E} f$ for every $C^2$ test function $f$ such that $f''$ is convex and $f$, $f'$, and $f''$ have polynomial growth. In particular, it gives an explicit upper bound for every absolute moment of order $p\ge3$. We then prove that, for every $p\ge4$, this bound is maximized by the centered square $g^2-1$ of a single standard Gaussian variable $g$. The resulting sharp inequality is \[ \bigl\|G^{\mathsf T}MG-\operatorname{tr}M\bigr\|_p \le \|g^2-1\|_p\,\|M\|_{\mathrm F}, \] with equality if and only if $M$ has rank one.
Read the full abstract

Computation and Language

Neuron-Guided Fine-Tuning: Unlocking Efficient Alignment Mechanisms for Large Language Models

Existing Supervised Fine-Tuning paradigms, particularly Full Parameter Fine-Tuning are often plagued by parameter redundancy, inconsistent data quality, and catastrophic forgetting, which current methods typically address in isolation and lack a unified optimization signal to bridge data selection, parameter updates, and knowledge preservation. To address this, we propose Neuron-Guided Fine-Tuning (NGFT), a holistic framework that leverages neuron activation patterns as a universal proxy to unify the fine-tuning lifecycle. NGFT operates via three synergistic mechanisms: (1) Adaptive Task-Specific Neuron Selection, which identifies essential neurons in a single forward pass to concentrate updates and reduce redundancy; (2) Activation-Based Data Selection, which prioritizes information-dense samples that maximize contribution to key neurons; and (3) Neuron Activation Alignment, a novel loss function that anchors activations to pre-trained states, deepening representation learning and preserving general knowledge. Experimental results across three models across both domain-specific and general benchmarks demonstrate that NGFT significantly outperforms existing mainstream fine-tuning methods in both efficiency and performance, while effectively mitigating catastrophic forgetting.
Read the full abstract

All authors Zeyu Wu, Junchao Wu, Shudong Liu, Runzhe Zhan, Xin Chen, Shu Yang, Yichao Du, Longyue Wang, Weihua Luo, Jinsong Su, Derek F. Wong

Computer Vision

FreeTransformSR: Efficient Lightweight Image Super-Resolution via Free Low-Rank Learnable Transform

Single image super-resolution aims to reconstruct high-resolution images from low-resolution inputs. This paper proposes FreeTransformSR, a novel lightweight super-resolution network based on a channel-wise free low-rank learnable transform. The transform learns task-adaptive basis functions in a data-driven manner, enabling adaptive feature modulation with minimal parameter overhead. To further enhance high-frequency detail recovery, we in-troduce a local feature modulation branch that complements transform-domain processing with depthwise convolu-tion. In addition, a soft complexity adaptive module dynamically fuses the outputs of local convolution and window self-attention branches through a lightweight gating network, adaptively adjusting the fusion ratio based on regional texture characteristics. An adaptive intensity modulation strategy is also incorporated to adjust transform-domain response strength at the sample level, enabling the network to dynamically adjust processing intensity according to input features. Extensive experiments on five benchmark datasets demonstrate that FreeTransformSR achieves competitive PSNR/SSIM performance with significantly fewer parameters and FLOPs. Specifically, FreeTransformSR achieves 32.41 dB on BSD100 (X2, 634K parameters) and 27.00 dB on Urban100 (X4, 653K parameters), while delivering faster inference speed than competing methods, making it well-suited for deployment in resource-constrained scenarios.
Read the full abstract

All authors Hongji Li, Yunhui Li, Jianing Wang, Xiaodong Wan

Cryptography and Security

Structurally Close, Temporally Distant: Measuring Security Exposure in Long-Horizon LLM Agents

Long-horizon LLM agents interact with untrusted content, persistent memory, external state, and sensitive tools. Existing analyses often characterize attacks by the number of execution steps between malicious input and a downstream action. We show that temporal remoteness can overstate security separation in stateful agents. We introduce a provenance-aware execution graph linking agent events through deterministic state, identifier, and tool provenance, and define influence distance $\DI$ as the shortest structural path from an untrusted source to a sensitive action. We compare it with sequence distance $\DT$, the shortest injection–sink path in the ordered trajectory. Since the influence graph contains every sequence edge, $\DI \leq \DT$; $\Gap=\DT-\DI$ measures the separation hidden by step count. Across 454 injection–sink pairs from 360 long-horizon AgentDojo trajectories over OpenAI's gpt-4o-mini and gpt-4o and Claude's Haiku 4.5 and Sonnet 4.6, $\Gap>0$ for 96.9% of pairs, with a median gap of 9 hops; 91.0% remain decoupled after removing the largest provenance-only edge class. On AgentDojo's banking suite, 33.8% of 231 pairs from 377 trajectories decouple through different provenance mechanisms. Among 274 OpenAI pairs, $\Gap$ does not independently predict attack success after controlling for $\DT$, attack family, and backend ($\beta_{\Gap}=0.066$, $p=.088$). At matched thresholds $k=2,3$, a deterministic $\DI$-based pre-execution gate blocks five attack sinks missed by a sequence-only gate with no additional benign blocking, although the paired gain is not significant ($p=.0625$). Execution structure therefore reveals proximity hidden by step count and can support targeted runtime intervention. We measure candidate influence pathways rather than causal attribution.
Read the full abstract

Computation and Language

UniRRM: Unified Reasoning Reward Models Across Languages and Evaluation Paradigms

Reinforcement learning (RL) excels on tasks with verifiable rewards, but in open-ended tasks, the reliability of reward models remains a key challenge. Existing solutions either depend on costly proprietary LLM-as-a-Judge systems or opaque scalar reward models that lack interpretability. Recent works on generative reward models offer a promising alternative, but they remain constrained by static evaluation criteria, fragmented evaluation paradigms, and limited multilingual support. To address these challenges, we introduce MixReward, a large-scale multilingual dataset spanning six domains and 103 languages, containing both pairwise and listwise data, and propose UniRRM, a unified reasoning reward model supporting multiple languages and evaluation paradigms. UniRRM uses a staged reasoning chain to dynamically generate task-generic and instruction-specific criteria, enabling fine-grained, input-adaptive judgments while maintaining consistency across languages. Experiments demonstrate that UniRRM-8B and UniRRM-14B achieve performance close to the state-of-the-art for models of comparable size across multiple benchmarks, and are effective for unseen evaluation paradigms. In addition, ablation studies validate the reliability and effectiveness of UniRRM.
Read the full abstract

All authors Peng Lai, Yichao Du, Junchao Wu, Weibo Gao, Linan Yue, Longyue Wang, Weihua Luo, Derek F. Wong, Guanhua Chen

Information Retrieval

Do All Nodes Benefit Equally from Knowledge Graphs? Adaptive Node-Aware KG Fusion for Recommendation

KG-aware recommendation has been widely studied to alleviate data sparsity by using knowledge graphs (KGs), which represent items, entities, and their relations as graphs and provide item-side knowledge. However, existing methods incorporate item knowledge without considering how much each user or item node should rely on it. As a result, they apply KG signals indiscriminately across nodes, even to nodes whose collaborative filtering (CF) signals from the interaction graph (IG) are already reliable. In this paper, we propose AdaKG (Adaptive Node-Aware KG Fusion), a novel KG-aware recommendation method that adaptively adjusts the contribution of auxiliary knowledge for each node. Since user-item interactions and item knowledge provide different types of signals, directly mixing them can distort the CF signals. To avoid this, AdaKG separately encodes the IG and KG with view-specific encoders, allowing each view to capture its own information. It then estimates how strongly each node should rely on item knowledge by measuring the stability of its CF signals under small adversarial perturbations, assigning a larger KG contribution to less stable nodes. Finally, AdaKG adaptively aligns the IG and KG embeddings in a shared space and fuses them according to the estimated node-wise reliance. Through experiments, we show that AdaKG achieves strong performance compared with its baselines and the effectiveness of our adaptive fusion strategy.
Read the full abstract

All authors Jaehyun Park, Minseo Jeon, Daewon Gwak, Sunuk Kim, Hanvit Lee, Jinhong Jung