Peer Revenue Play credits. Real papers.

Papers

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

All of arXiv

Friday 9 October

Machine Learning

S$^3$R: Fast and Faithful RL Rollouts via Self-Sparse Speculative Decoding

Yuhang He, Hongyu Ge, Zihan Lin et al. (+2 more)

Fast rollout generation for long-context reinforcement learning with verifiable rewards (RLVR) requires reducing KV-cache traffic without changing the response groups used for policy optimization. In our profiled RLVR workloads, attention accounts for 72–76% of autoregressive decoding kernel time. Sparse attention reduces these reads, but sampling directly from a sparse policy changes the response groups used for learning. For group-relative optimization, we show that even exact response-wise reweighting does not generally recover the expected dense-policy update because the advantage of each response depends on the rewards of other group members. We propose Self-Sparse Speculative Rollouts (S$^3$R). It uses sparse attention to propose tokens, then generates responses through full-attention verification and exact speculative sampling. Drafting and verification share the same policy parameter, eliminating the need to train or update a separate drafter. The resulting response groups follow the dense-policy distribution. Under S$^3$R, we characterize speedup conditions and the costs of drafting and verification. We found that dense verification still requires the complete historical KV cache. A KV-aware scheduler therefore limits KV-cache repeated offloading and restoration while maintaining rollout parallelism. % OR thrashing In a matched ablation with an aggressive sparse drafting budget, S$^3$R maintains stable training while direct sparse sampling with training-side correction becomes unstable. On long-context mathematical reasoning, S$^3$R achieves $1.39\times~1.35\times$ the end-to-end training throughput of autoregressive rollouts on Qwen3 model, with comparable task performance.
Read the full abstract

All authors Yuhang He, Hongyu Ge, Zihan Lin, Siyi Liu, Yongqi Zhang

Logic in Computer Science

Compact Quantitative Equational Theories as Monad Restrictions

Matteo Mio and Colin Riba

Proofs in quantitative algebra are, generally, infinite objects due to the presence of infinitary rules in the deductive apparatus. A quantitative equational theory is called compact if all its consequences are derivable by means of finite proofs, that is, proofs not using any instances of the infinitary rules. This definition is purely logical. In this work we give an equivalent, but categorical, characterisation based on the notion of monad restriction.
Read the full abstract

Earth and Planetary Astrophysics

Detailed study of HD 219134: a benchmark K-dwarf with a compact planetary system

Jérôme Bétrisey, Camilla Pezzotti, Gaël Buldgen et al. (+4 more)

Space missions such as CoRoT, Kepler, and TESS have transformed asteroseismology and its synergies with exoplanet science. We revisit the nearby K-dwarf HD 219134 to investigate the tension between asteroseismic and interferometric radii, update its fundamental parameters, and reconstruct the rotational, magnetic, and atmospheric evolution of the system. We combine forward asteroseismic modeling with seismic inversions to assess systematics from non-seismic constraints, radius measurements, and surface modeling. The revised stellar properties are then used for rotational evolution, stellar-wind, and high-energy irradiation calculations. Seismic inversions rule out the most recent interferometric radius, while the asteroseismic and spectroscopic results are compatible with the 2019 interferometric measurement. We obtain $M=0.779\pm0.015_\mathrm{(stat)}\pm0.006_\mathrm{(sys)}\ M_\odot$, $R=0.753\pm0.005_\mathrm{(stat)}\pm0.002_\mathrm{(sys)}\ R_\odot$, and an age of $8.92\pm0.89_\mathrm{(stat)}\pm0.19_\mathrm{(sys)}$ Gyr. These systematics include only non-seismic constraints; literature combined uncertainties from physical inputs and surface effects are $0.007M_\odot$, $0.002R_\odot$, and $0.81$ Gyr. Rotational and wind modelling shows that standard magnetic braking is broadly compatible with the present-day rotation and angular-momentum loss. Weakened braking gives a closer match but challenges standard wind-loss scalings. The inferred X-ray history suggests early stripping of the innermost planets, while the outer planets likely retained part of their primordial envelopes; HD 219134d experienced only moderate erosion. The apparent radius discrepancy likely reflects underestimated uncertainties in previous interferometric analyses rather than an astrophysical inconsistency. We advocate adopting the asteroseismic radius, which is robust to modeling assumptions and optimisation strategies.
Read the full abstract

All authors Jérôme Bétrisey, Camilla Pezzotti, Gaël Buldgen, Aline A. Vidotto, Colin P. Folsom, Michaël De Becker, Michaël Gillon

Risk Management

Enhancing credit risk scorecards with transactional-based behavioural features

Dian Prinsloo, Tanja Verster and Neill Smit

Background: Credit risk modelling is fundamental in financial institutions, typically using credit history, repayment behaviour, income, and outstanding debt to estimate a borrower's probability of default. Although transactional data are sometimes incorporated through aggregated measures such as recency, frequency, and monetary indicators, behavioural information embedded in raw transaction narratives remains largely unexplored due to its unstructured nature. This limits opportunities to extract richer behavioural insights. Aim: This research investigates payment narratives from transactional data as an alternative data source for credit risk modelling. Setting: The study uses transactional data from a large South African financial institution focused on small and medium enterprises. Method: Techniques such as natural language processing and regular expressions are used to transform unstructured transaction descriptions into meaningful, explainable behavioural features by categorising transactions into interpretable spending categories. A novel categorisation algorithm was developed and evaluated for classification accuracy. The extracted features are incorporated into a credit risk scorecard to assess their predictive value. Results: The categorisation algorithm demonstrates high classification accuracy. Using these features in the scorecard resulted in a significant increase in Gini compared to a scorecard without them. Conclusion: Text-based transactional information can be systematically extracted, categorised, and operationalised to enhance credit risk model performance. Contribution: The study develops a novel, integrated, explainable methodology for transforming transactional text into behavioural features, introduces a structured RegEx-based categorisation approach, and demonstrates their predictive value within traditional scorecard frameworks.
Read the full abstract

Sound

QC-MI-OVSEL: Benchmarking Query-Conditioned Multi-Instance Open-Vocabulary Sound Event Localization

Yongqiang Chen, Peiwei Chang, Yuelan Cheng et al. (+2 more)

We study query-conditioned multi-instance open-vocabulary sound event localization (QC-MI-OVSEL), where a system must recover all active physical sources matching a natural-language category query, including when the queried category is absent. We introduce the QC-MI-OVSEL benchmark, a 40,000-scene first-order Ambisonics benchmark with 80,000 balanced present/absent query views over 83 classes, including 17 novel classes excluded from training and development. Each frame is annotated with the activity and 3-D direction of every query-matching source, with up to three simultaneous targets. We further propose Q-RSet, which uses query-conditioned source slots and responsibility-based evidence aggregation for source-set prediction. Experiments show improved source recovery, localization recall, cardinality estimation, exact-set recovery, and absent-query suppression over complementary baselines. Analysis indicates that novel categories, paraphrased queries, and higher source multiplicity mainly impair source discovery and set completion, while directional accuracy remains comparatively stable for successfully recovered sources.
Read the full abstract

All authors Yongqiang Chen, Peiwei Chang, Yuelan Cheng, Xiatian Zhu, Wenwu Wang

Machine Learning

Does Learning Recurrent Networks Matter in Partially Observable RL?

Zekang Wang, Zhe He and Gaojie Jin

Reinforcement learning (RL) under partial observability typically re- lies on recurrent networks to encode temporal information, with the recurrent module trained jointly with the policy or value function. This raises a natural question: is learning the recurrent network actually necessary? In this paper, we investigate whether a frozen recurrent network can provide sufficient memory for solving par- tially observable Markov decision processes (POMDPs). Motivated by the theoretical foundations of Reservoir Computing (RC) and Generalized Reservoir Computing (GRC), we study a simple archi- tecture that combines a fixed recurrent network with a trainable nonlinear output module. We evaluate this approach across tasks with varying memory demands, observation modalities, and action spaces. Surprisingly, frozen recurrent networks support recall over hundreds of steps and achieve competitive control performance across diverse POMDPs, while using less than half the number of trainable parameters required by their fully trainable counterparts. These results suggest that learning the recurrent memory module is not always necessary for solving POMDPs. Frozen recurrent networks therefore provide a simple, parameter-efficient, and sur- prisingly strong baseline for partially observable RL. Our source code is available at https://github.com/Aphlios/FRN-POMDP
Read the full abstract

Mesoscale and Nanoscale Physics

Graphene double quantum dot as a solid-state two-level system in the THz range

Alexis Wietzke, Solen Coeymans, Benjamin Hue et al. (+12 more)

Matter-based two-level quantum systems are fundamental building blocks of quantum technologies, underpinning applications in quantum computing, simulation, communication, and sensing. Today, most of these two-level quantum systems, including Rydberg atoms, trapped ions, superconducting circuits, molecules, and quantum dots, involve transition frequencies in microwave or optical domains. However, extending their transition frequency to the THz spectral range could provide numerous benefits, leveraging the unique properties of THz radiation. Here, we show that an hBN-encapsulated graphene double quantum dot realizes a two-level system with a transition frequency in the THz spectral range. Using low-temperature transport spectroscopy in the dark combined with modeling based on nonequilibrium Green's function formalism, we demonstrate the formation of an artificial molecule, featuring bonding and antibonding states separated by an energy of up to 0.6 meV, corresponding to a resonance frequency of 0.145 THz. This approach provides direct access to important parameters of open double quantum dot systems, including tunneling rates, offering a practical alternative to conventional techniques such as photon-assisted tunneling, which are challenging to implement at THz frequencies. The realized 2D-material-based THz two-level system, compatible with direct on-chip integration, could serve as a key building block for future THz quantum technologies. Furthermore, by involving a single-electron transition, coupling these devices to a THz resonator could enable the exploration of ultra-strong light-matter interactions at the fundamental limit of a single transition coupled to a single cavity, beyond the reach of existing THz quantum systems based on collective transitions.
Read the full abstract

All authors Alexis Wietzke, Solen Coeymans, Benjamin Hue, Simon Messelot, Aurelien Schmitt, Michael Rosticher, Kenji Watanabe, Takashi Taniguchi, Jerome Tignon, Sukhdeep Dhillon, Rebeca Ribeiro, Audrey Cottet, Matthieu Delbecq, Takis Kontos, Juliette Mangeney

Robotics

PATACON: GPU-Parallel Tangent-Bundle RRT-Connect for Constrained Motion Planning

Gahyun Oh, Junho An, Suhyun Jeon et al. (+1 more)

Fast whole-body motion planning enables continuous replanning for humanoid robots in dynamic environments while satisfying collision avoidance and kinematic constraints. However, constrained motion planning incurs substantial computational cost due to repeated constraint projection in addition to collision checking and nearest-neighbor search. Recently, GPU-based massively parallel planners have been proposed to reduce computational cost. Nevertheless, these planners still steer in the ambient configuration space, so projecting candidates that deviate from the manifold remains a significant source of overhead, particularly in highly constrained, high-dimensional problems. In this work, we introduce PATACON, a GPU-parallel bidirectional planner that explores tangent space to exploit the local manifold geometry induced by known constraint functions. Experiments on dual-arm closed-chain and humanoid whole-body constrained planning tasks show that PATACON outperforms the prior work as configuration-space dimension and constraint codimension grow, achieving up to an 11.25 times speedup. These results suggest that GPU-parallel tangent-space exploration brings constrained whole-body motion planning closer to real-time performance.
Read the full abstract

All authors Gahyun Oh, Junho An, Suhyun Jeon, Suhan Park

Artificial Intelligence

JANUS: Enhancing Long-Horizon Autonomous Research with Co-Evolving World Models

Mengdi Liu, Chenghao Jia, Hong Chang et al. (+3 more)

Long-horizon autonomous research has shown promise in accelerating scientific progress through sustained cycles of hypothesis generation, implementation, and validation. However, this Generate-Execute-Feedback paradigm relies on reactive trial-and-error, where costly and time-consuming executions introduce severe feedback delays, throttling evolutionary progress under explosively expanding search space. To address this challenge, we propose JANUS, a framework that unifies future outcomes forecasting and past experience reflection through a Predict-Verify-Evolve loop driven by a co-evolving world model. Within an evolving graph, the research agent proposes candidates while world model scores them by forecasting value improvement, information gain, and exploration utility, then accordingly selects, refines, or prunes branches prior to execution. After receiving ground-truth feedback, JANUS distills experience into evolving skills: hierarchical skills for research agent, exploration-strategy skills for world model, and their coordination protocols. This mechanism enables continual co-evolution: the research agent sharpens candidate generation while the world model refines its predictive judgment over extended horizons. Extensive evaluations on MLS-Bench-Lite and real research tasks demonstrate that JANUS significantly outperforms state-of-the-art baselines, achieving both higher evolving efficiency and ceiling.
Read the full abstract

All authors Mengdi Liu, Chenghao Jia, Hong Chang, Mingjie He, Shiguang Shan, Xilin Chen

Mesoscale and Nanoscale Physics

Equal-Spin local and nonlocal Andreev reflection in altermagnet/superconductor/altermagnet junction

Kai-Yi Lyu, Xing Wang and Yu-Xian Li

We investigated the equal-spin Andreev reflection phenomenon in an altermagnet/super\-conductor/altermagnet (AM/SC/AM) junction, focusing on the roles of the crystal orientation angle and the Néel vector. The crystal orientation angle in the AM effectively determines the spin-dependent transport channels for both electrons and holes, leading to orientation-dependent local and crossed Andreev reflection in the AM/SC/AM junction. We studied how and under what conditions this equal-spin Andreev reflection occurs. Rotating the Néel vector in the AM substantially increases spin-flip channels for both electrons and holes, giving rise to equal-spin Andreev reflection. When the Néel vector $\theta=\pi/2$, the spin-dependent transmission channels for crystal orientations $\alpha=0$ and $\alpha=\pi/2$ are completely the same, the Andreev reflection coefficients remain different, reflecting the distinct roles of the Néel vector in spin space and the crystal orientation in real space. By tuning the crystal orientation angle and the Néel vector, orientation-dependent conventional and equal-spin Andreev reflection can be realized. This study reveals that Néel vector rotation is the direct cause of equal-spin Andreev reflection in AM/SC/AM junctions and provides a viable pathway for realizing spin-dependent superconducting transport phenomena in AM heterojunctions.
Read the full abstract

Algebraic Geometry

Classification of can-Var diagrams of mixed Hodge structures and monodromic mixed Hodge modules on $\mathbb{C}$

Takahiro Saito

By our previous work, the category of unipotent monodromic mixed Hodge modules on $\mathbb{C}$ is equivalent to the category of can-Var diagrams of mixed Hodge structures. In this paper, we show that the associated graded diagram with respect to the weight filtrations decomposes as a finite direct sum of four basic types tensored with pure Hodge structures, which we call a split graded model. We then show that every unipotent monodromic mixed Hodge module is represented by a split graded model together with a finite family of $\mathbb{C}$-linear maps. We also obtain an analogous description in the case where the monodromy has no eigenvalue $1$.
Read the full abstract

High Energy Physics (Theory)

Dirac-Born-Infeld action from inequivalent representations of open strings

Isao Kishimoto, Tomohiko Takahashi and Ayane Tezuka

We study open strings in a constant magnetic field from the viewpoint of universal canonical variables. Using a conserved Klein-Gordon-type pairing, we show that the canonical commutation relations remain background independent in the distributional sense, while the background dependence is encoded in the mode representation. The Bogoliubov transformation between different magnetic backgrounds is not unitarily implementable, so that the corresponding Fock representations are inequivalent. Nevertheless, the transformation determines the relative normalization of the vacua. We show that the resulting relative vacuum normalization precisely reproduces the relative boundary $g$-function, or equivalently the magnetic-field-dependent part of the disk amplitude, and yields the Dirac-Born-Infeld factor. Thus the disk normalization is encoded directly in the transformation between inequivalent representations of universal open-string variables.
Read the full abstract

Robotics

Prediction-Informed Landing Window Planning for UAV Landing on USV in High Sea States

Kaizhen Li, Xi Zhou, Dan Zhang et al. (+1 more)

Autonomous landing of unmanned aerial vehicle (UAV) on unmanned surface vehicle (USV) is difficult in high sea states. Existing hierarchical landing methods often infer future USV motion only implicitly from short observation histories, which limits their ability to adapt to severe and rapidly changing sea conditions. This paper presents a prediction-informed hierarchical control framework. A pretrained model is developed to explicitly predict the USV attitude during high-level landing decisions. The prediction helps identify suitable landing windows and choose the right touchdown time, while a conventional low-level flight controller maintains UAV stability and tracks motion commands. To evaluate the framework under realistic disturbances, a real-sim validation setup is constructed in which USV-motion data collected in real sea conditions are transferred to a MuJoCo USV. The experiments demonstrate the feasibility of using explicit attitude prediction to improve landing-window selection and support timing-aware UAV landing planning in rapidly varying marine environments.
Read the full abstract

All authors Kaizhen Li, Xi Zhou, Dan Zhang, Xiaowei Li

Materials Science

Insights on the volume variation of LiCoxMn2-xO4 electrode upon cycling: a correlative operando XRD and XAS study

Eunike Mahayoni, Louiza Larbi, Romain Batiot et al. (+4 more)

Understanding volume changes in high-voltage spinel cathodes is essential for mitigating mechanical degradation and improving cycling stability. However, linking these changes to cation-specific redox processes remains challenging. Here, we combine operando X-ray diffraction (XRD) and X-ray absorption spectroscopy (XAS) to investigate the LiCoxMn2-xO4 spinel system, which exhibits a cubic structure and isotropic lattice response, in order to correlate the evolution of the local and long-range structure. Five compositions (x = 0, 0.25, 0.5, 0.75, and 1.0) were synthesized and electrochemically characterized, with x = 0, 0.5, and 1.0 examined in detail under identical cycling conditions. XAS at the Mn and Co K-edges reveals distinct Mn4+/Mn3+ and Co4+/Co3+ redox processes, with Mn redox activity dominating below 4.7 V and Co redox triggered at higher potentials. Chemometric tools for XAS data analysis further resolve the evolution of local structure during cycling. Both XRD and XAS capture consistent trends in volume evolution, despite probing different structural length scales. Increasing Co content progressively decreases the overall volume change, from 5.0 % to 1.6 % by XAS and from 5.8 % to 2.1 % by XRD. These results demonstrate that Co substitution suppresses electrochemically induced structural strain and establish a direct link between cation-specific redox activity and volumetric evolution. The correlative operando XRD/XAS approach provides a framework to investigate newly designed low-strain, high-energy cathodes with improved cycling stability.
Read the full abstract

All authors Eunike Mahayoni, Louiza Larbi, Romain Batiot, Anatole Weill, Damien Brégiroux, Antonella Iadecola, Arnaud Perez

Human-Computer Interaction

olBlend: Language-Guided Molecular Blend Retrieval with Mixture Odor Prediction

Dian Jin, Zhechang Chen, Qilin Liang et al. (+5 more)

Converting open-ended natural-language odor descriptions into molecular mixtures that can be physically prepared from an available inventory remains an important challenge in computational olfaction. We propose olBlend, a language-guided framework for molecular mixture retrieval. olBlend first uses MixPOM to predict odor probabilities for candidate molecular pairs and stores these predictions for retrieval. It then uses a language model to translate each odor description into three types of semantic requirements: lead attributes that define the main odor, ambient attributes that describe its background, and absent attributes that should be avoided. A structured scoring algorithm uses the precomputed odor probabilities to score how well each candidate mixture meets these requirements, then retrieves and ranks mixtures from the available inventory. Evaluation on established odor datasets shows that MixPOM improves mixture odor prediction over averaged single-molecule predictions. Retrieval comparisons across natural-language queries demonstrate how background and exclusion requirements guide mixture selection. We further prepare retrieved formulations as physical odor samples, demonstrating the workflow from natural-language descriptions to realizable molecular formulations.
Read the full abstract

All authors Dian Jin, Zhechang Chen, Qilin Liang, Yonghui Luo, Zixun Fu, Defeng Sun, Yancheng Yuan, Xiaoming Tao

Computer Vision

Destination-Conditioned Adaptation with Semantic-Geometric Alignment for Generative Inbetweening

Zhiyu Xu, Wei Shang, Yushen Xu et al. (+4 more)

Generative inbetweening aims to synthesize temporally coherent intermediate frames between a start frame and a target frame. Although pretrained image-to-video diffusion models provide strong priors for large-motion inbetweening, they are not explicitly designed to reach a specified target frame. They may therefore generate realistic individual frames while drifting from the target or producing unstable motion under large displacements and non-rigid dynamics. We propose DSG-GI, a framework that adapts pretrained latent video diffusion models through Destination-Conditioned Adaptation and Semantic-Geometric Alignment. Destination-Conditioned Adaptation treats the target frame as the destination of the sequence and provides temporally progressive guidance along the generated trajectory. Semantic-Geometric Alignment aligns the token-wise spatio-temporal relational geometry of backbone features with that of a frozen video foundation encoder and applies a flow-matching-aware latent flow constraint to samples near the data endpoint of the probability path. Together, these components strengthen destination guidance and preserve coherent semantic-geometric motion structure. Extensive experiments on standard and large-motion inbetweening benchmarks demonstrate improvements in temporal stability, target-frame faithfulness, and perceptual quality. Our code will be made publicly available.
Read the full abstract

All authors Zhiyu Xu, Wei Shang, Yushen Xu, Wenwen Li, Zhiyan Xu, Pengfei Zhu, Dongwei Ren

Machine Learning

WaSA: Wavelet Scattering Autoencoders for Unsupervised Offline Detection of Malicious Clients in Federated Learning

Alessandro Licciardi

Federated Learning (FL) enables collaborative model training without exposing raw data, yet remains vulnerable to malicious clients injecting poisoned datasets. Existing defenses are typically reactive (operating during training) or supervised (requiring prior knowledge of attack types) - assumptions that rarely hold in open-ended IoT deployments. We propose WaSA, a fully unsupervised, offline framework for malicious client detection. Before training, each client transmits a single, privacy-preserving spectral summary of its data to the server. A denoising autoencoder, trained solely on clean auxiliary data, scores clients based on their deviation from the learned benign manifold, identifying corruptions regardless of the attack type. WaSA requires no attack labels, generalizes to unseen anomalies, and introduces a Boltzmann sampling strategy for client selection that smoothly interpolates between hard exclusion and standard aggregation. Experiments on MNIST, CIFAR-10, and CIFAR-100 under diverse attack scenarios (up to 90 % malicious clients) demonstrate consistent improvements over supervised baselines and robust aggregation methods, with negligible communication overhead. Code is available at https://github.com/alelicciardi99/wavelet_autoencoders_malicious.
Read the full abstract

Human-Computer Interaction

From Observation to Intervention: Identifiability of Latent User Bias in Human-AI Interaction

Lucas Biechy, Yuxiao Li, Zhonghao He et al. (+1 more)

Alignment methods typically treat user behavior as a reliable proxy for underlying preferences. However, this assumption breaks down when users hold belief-dependent cognitive biases, risking irreversible “confirmation traps” and harmful feedback loops. In this paper, we formalize when an AI system can distinguish true user preferences from behavior shaped by cognitive bias. Modeling this interaction as a multi-armed bandit, we first prove a non-identifiability theorem: under passive observation, a static binary biased user produces action trajectory distributions strictly identical to an unbiased user in a transformed environment, forcing any action-only estimator to perform no better than random guessing. We then show that active preference elicitation breaks this observational equivalence. Specifically, we prove that an Observer applying a finite sequence of adversarial zero-reward interventions of length $m_\delta$ makes the post-intervention action distributions arbitrarily separable. Our theoretical results suggest that robust alignment requires actively intervening to tell informed preferences from biases.
Read the full abstract

All authors Lucas Biechy, Yuxiao Li, Zhonghao He, Tianyi Qiu

Artificial Intelligence

AdaC-GRPO: Towards Mastering Long-Horizon Visual Sudoku Puzzles via Adaptive Curriculum Learning

Sida Li, Yuanxin Liu, Linli Yao et al. (+4 more)

While reinforcement learning with verifiable reward (RLVR) enhances reasoning in large language models, its efficacy in long-horizon multimodal tasks remains underexplored. Visual Sudoku puzzles pose a representative challenge, requiring multi-step deductive reasoning over both local constraints and global dependency chains. To this end, we introduce FG-Sudoku, a fine-grained benchmark categorized by grid size and required deductive steps. Evaluations reveal that frontier MLLMs fail on complex long-range dependencies (e.g., 9*9 puzzles). During post-training, this near-zero success rate causes severe reward sparsity, halting standard RLVR. We propose Adaptive Curriculum Group Relative Policy Optimization (AdaC-GRPO) to overcome this. Unlike rigid or prior-based schedules, AdaC-GRPO dynamically constructs training batches by tracking the model's "Zone of Proximal Development" via online rollout accuracy. Experiments demonstrate that AdaC-GRPO enables Qwen2.5-VL-7B to approach the performance of Gemini-2.5-Flash while bridging 62.2% of the overall accuracy gap to Gemini-2.5-Pro, with its generalizability further validated across InternVL-3.5 and Qwen3-VL.
Read the full abstract

All authors Sida Li, Yuanxin Liu, Linli Yao, Shicheng Li, Xu Sun, Haoli Bai, Xiaohui Li

High Energy Astrophysical Phenomena

Simulation Study of Reconstruction of Cosmic Ray Extensive Air Shower with Thermal Neutron Detected by ENDA

Fan-Ping Li, Xin-Hua Ma, Hu Liu et al. (+25 more)

In the investigation of the "knee" region within the cosmic ray energy spectrum, extensive air shower (EAS) reconstruction by using different EAS secondary particles are of paramount importance for energy determination and composition separation and finally measurement of energy spectrum of different cosmic ray composition. The Electron-Neutron Detector Array (ENDA) is designed to detect thermal neutrons generated in the ground by high energy hadrons which constitute the "backbone" of EAS and carry vital information regarding the primary cosmic rays. At present, ENDA-64, consisting of 64 detectors, is employed inside the Large High Altitude Air Shower Observatory (LHAASO). This work simulates the response of ENDA and investigates the lateral distributions of thermal neutrons. A comparison of the NKG and double-exponential functions for describing this distribution using MC simulation data shows that the double-exponential function is more accurate. The mean difference between fitted and measured thermal neutron density is within ~ 6%. Meanwhile, the thermal neutron density at 10 m from the shower axis, $\rho_n$, depends mainly on the sum of all hadrons energy or the number of high-energy hadrons. The relationship between $\rho_n$ and EAS hadron energy or the number of high-energy hadrons can be well fitted by a linear function on a log scale. This work is fundamental for energy determination and composition separation with hybrid detection by using ENDA in the near future.
Read the full abstract

All authors Fan-Ping Li, Xin-Hua Ma, Hu Liu, Tian-Lu Chen, Shu-Wang Cui, Danzengluobu, Xin-Rui Gao, Wei Gao, Wen-Chao Gao, Hai-Bing Hu, Denis Kuleshov, Kirill Kurinov, Bing-Bing Li, Mao-Yuan Liu, Ye Liu, Igor Maliy, Da-Yu Peng, Yao-Hui Qi, Jia-Bo Qin, Dong Qu, Oleg Shchegolev, Yuri Stenkin, Yu-Sen Wang, Tian-Shuang Yang, Li-Qiao Yin, Hui-Qian Zhang, Liang-Wei Zhang, Shi-Yuan Zhang

High Energy Astrophysical Phenomena

Flux Evolution of PSR B0540–69 and the Spectral Energy Distribution of its Pulsar Wind Nebula

Juan Zhang, Xuezhi Liu, Junyang Zhang et al. (+12 more)

The young Crab-like pulsar PSR B0540–69, located in the Large Magellanic Cloud, underwent a spin-down rate transition (SRT) in 2011 December, after which its pulsar wind nebula (PWN) was reported to brighten up to 2018. In this work, we extend the study of its X-ray and $\gamma$-ray flux evolution through 2026 by combining observations from multiple X-ray instruments with data from the Fermi Large Area Telescope, thereby constructing a flux history of the PSR$+$PWN system spanning more than two decades. We find that the enhanced X-ray luminosity of the PSR$+$PWN system has persisted for more than 14 yr after the SRT, up to early 2026, whereas the pulsed X-ray flux remains unchanged across all epochs. This confirms that the brightening originates from the PWN. In the $\gamma$-ray band, neither the pulsed nor the non-pulsed flux shows any detectable long-term variation between 2008 and 2025. Together, these results imply that the additional spin-down energy released at the SRT has been continuously injected into the PWN over decades, placing strong constraints on the energy budget and physical mechanism of the transition. Combining the non-pulsed X-ray and $\gamma$-ray fluxes measured here with previously published radio-to-UV data, we further construct a broadband spectral energy distribution of the PWN, which shows properties markedly different from those of the Crab Nebula.
Read the full abstract

All authors Juan Zhang, Xuezhi Liu, Junyang Zhang, Hanlong Peng, Mingyu Ge, Songzhan Chen, Shanshan Weng, Shijie Zheng, Shumei Jia, Chengkui Li, Haisheng Zhao, Ju Guan, Yong Chen, Liming Song, Fangjun Lu

Computer Vision

MixTok: Adaptive Visual Tokenization with Global 1D Tokens and Sparse 2D Refinement

Chenge Li, Yefei He, Xinyu Liu et al. (+3 more)

Efficient visual tokenization requires balancing representation compactness with the preservation of fine-grained image content. Conventional grid-based tokenizers allocate tokens uniformly across spatial locations despite non-uniform visual complexity, whereas compact 1D tokenizers aggregate global information but can struggle to preserve local details. We introduce MixTok, a hybrid visual tokenizer that combines a compact 1D global representation with sparse 2D tokens for local refinement. A content-adaptive router ranks candidate locations by refinement priority and determines the number of 2D tokens for each image under a target average token budget. At the highest-scoring locations, 2D token features replace the spatial features recovered from the 1D representation, and a shared decoder reconstructs the image from the resulting mixed features. On ImageNet 256 x 256, MixTok achieves highly competitive reconstruction and generation performance, with an rFID of 1.195 and a gFID of 1.95 at respective average budgets of approximately 128 and 131 content tokens per image.
Read the full abstract

All authors Chenge Li, Yefei He, Xinyu Liu, Feng Chen, Hong Zhou, Bohan Zhuang

Spectral Theory

On the inequality $\mu_3 < \lambda_1$ for the eigenvalues of the Laplace operator

N. Filonov

In 1952, Pólya proved the inequality $\mu_2 < \lambda_1$ for the eigenvalues of Dirichlet and Neumann problems in domain $\Omega \subset \mathbb{R}^d$, $d \ge 2$. In 2026, Tang and Zhang proved the inequality $\mu_3 < \lambda_1$ in the two-dimensional case. We show that $\mu_3 < \lambda_1$ for $d\ge 3$ as well.
Read the full abstract

Computer Vision

High-Throughput Generative Video Super-Resolution on a Single Consumer GPU

Georgios Panaretou and Sotirios Chatzis

Generative video super-resolution (VSR) can recover visually compelling high-frequency detail by leveraging diffusion priors, but iterative sampling and heavy spatio-temporal coupling make practical deployments difficult. We study a different operating point: practical HD generative VSR on a single high-end consumer GPU without end-to-end retraining. To this end, we present a zero-shot framework built entirely from pretrained components. The proposed system decouples spatial enhancement from temporal stabilization. A time-distilled one-step diffusion backbone handles spatial upscaling, while temporal coherence is promoted by a lightweight pretrained flow-based interpolator that synthesizes intermediate frames between super-resolved keyframes. To further improve scalability, we introduce a resolution-aware hybrid attention scheme that, in a zero-shot manner, replaces selected high-resolution global-attention layers with neighborhood attention while preserving global attention in semantically critical low-resolution layers, while also improving temporal consistency. Experiments show that the proposed framework reaches up to ~32 FPS at 720p resolution under interleaved inference on a single NVIDIA RTX 4090. At 4K, the proposed attention substitution can yield up to a 44% speedup over the original attention implementation. Across standard VSR benchmarks, the method consistently shifts the speed-quality trade-off decisively toward runtime efficiency, trading reconstruction fidelity for strong no-reference perceptual quality, while remaining reasonably competitive against generative approaches in temporal consistency. It also exposes clear limitations in handling more challenging motion, which a simple motion gate partially mitigates. Code will be released upon acceptance.
Read the full abstract

Algebraic Geometry

Boundedness of polarized log Calabi-Yau fibrations

Haoran Zhu

We investigate the boundedness problem for log Calabi-Yau fibrations whose bases and general fibers are bounded. We improve the results in arXiv:2504.05243 from birational boundedness to actual boundedness, which confirms a conjecture of Birkar and Hacon when the boundary coefficients belong to a fixed finite rational set. We also get a new proof for the birational boundedness result in arXiv:2504.05243 which is substantially more conceptual.
Read the full abstract

Computer Vision

CompactSplat: Spatially Adaptive Gaussian Distribution for Feedforward Scene Reconstruction

Bing He, Jingnan Gao, Yunuo Chen et al. (+4 more)

Reconstructing 3D scenes from sparse views without per-scene optimization remains highly challenging, especially for recovering accurate geometry and fine textures. While recent feedforward approaches leverage generalizable 3D Gaussian Splatting (3DGS) for scene generation, they typically assign one or multiple Gaussians to each pixel. Such uniform allocation not only produces highly redundant representations by wasting primitives in homogeneous areas, but also fails to exploit the inherent geometric priors of planar surfaces, often resulting in sub-optimal structural modeling. To address these limitations, we present CompactSplat, a novel feedforward framework that enables a content-aware allocation of 3D Gaussians. By integrating texture-guided spatial partitioning with hierarchical sampling, CompactSplat adaptively distributes primitives - concentrating dense Gaussians in complex regions while tiling flat areas with expanded, scale-modulated primitives. Moreover, our framework inherently supports dynamic sparsity control at inference time, enabling seamless trade-offs between rendering fidelity and memory footprint without requiring network retraining. Extensive experiments on RealEstate10K, DL3DV, and ScanNet demonstrate that CompactSplat consistently outperforms prior methods on both standard metrics and high-resolution rendering consistency, achieving high-fidelity reconstructions with significantly fewer primitives.
Read the full abstract

All authors Bing He, Jingnan Gao, Yunuo Chen, Zhengxue Cheng, Rong Xie, Li Song, Wenjun Zhang

Computer Vision

Fix the Match, Not the Solver: STITCH – a Self-Supervised Compatibility Measure for Eroded Visual Puzzles

Nadav Alali and Ohad Ben-Shahar

Unlike the clean and pixel-aligned complete pieces of standard jigsaw benchmarks, real-world puzzle fragments lose their boundary regions to erosion, destroying the signal that classical compatibility measures (CMs) rely on. We first re-establish, via oracle experiments, that the CM, not the solver, is the performance bottleneck: with a perfect CM, even a simple greedy solver achieves perfect reconstruction. This motivates STITCH, a self-supervised CM that sidesteps missing boundaries by comparing frozen ViT patch tokens from piece interiors via a lightweight cross-attention head (~249K trainable parameters), trained without manual annotation. Plugged into unchanged compatibility-based solvers, STITCH reaches .992/.954 Abs/Perf on JPwLEG-5 (+.629 Perf over the best prior method), nearly saturating the benchmark. On a new ImageNet 7x7 benchmark, it outperforms all alternatives, including inpaint-then-match pipelines built on modern generative models. Importantly, performance remains robust across erosion levels and generalizes to puzzles with over 40% missing pieces without retraining. Trained on the recent GAP datasets of irregular, heavily eroded fragments, the same CM recipe enables compatibility-based solvers to surpass recent learning-based solvers on GAP-3 and GAP-5. Since this leap in performance is achieved by a pluggable CM module, the takeaway is clear: fix the match, not the solver.
Read the full abstract

Sound

How Should We Evaluate Naturalness for Synthetic Speech with a Foreign Accent?

Kentaro Onda, Satoru Fukayama, Daisuke Saito et al. (+1 more)

Recent advances in speech synthesis technology have enabled the generation of speech with diverse characteristics. One such characteristic is a foreign accent. In our preliminary experiments, we found that UTMOS, a widely used automatic speech quality assessment system, tended to assign lower scores to speech with stronger foreign accents, even though it is produced by a human speaker. This observation raised a question: How do humans evaluate the naturalness of foreign-accented speech? We conducted a large-scale subjective evaluation involving both native and non-native listeners under different evaluation instructions. The results showed that humans also considered accentedness when evaluating naturalness, but that defining naturalness in terms of human-likeness and selecting raters who have greater exposure to the target accent may enable more appropriate evaluation. This study aims to provide a starting point for future discussions on fairer and more appropriate evaluation of synthetic speech with a foreign accent.
Read the full abstract

All authors Kentaro Onda, Satoru Fukayama, Daisuke Saito, Nobuaki Minematsu

Computer Vision

PreCoG: Preview-Edit for Efficient Video Correction in Generation-Time

Yilai Liu, Wenqi Pei, Zhifeng Lu et al. (+3 more)

Iterative text-to-video generation relies on feedback to correct compositional and temporal failures. However, existing methods typically require another full generation run for each correction, repeating computation and potentially disrupting already correct content. We introduce PreCoG, an iterative framework for correcting semantic failures during video generation. PreQA diagnoses failures from low-cost previews and constructs corrective conditions, while PreEdit edits cached generation trajectories under these conditions, attenuating orthogonal changes to limit disruption of already correct content. PreCoG completes generation only for the selected candidate, reusing its cached trajectory to reduce redundant computation and help preserve semantic consistency between the preview and the final video. Experiments show that PreCoG achieves the highest VBench and T2V-CompBench scores and approaches VSR's leading VS2 score at $0.56\times$ its inference time. It also outperforms GENMAC on all three quality and alignment metrics at $0.86\times$ its inference time.
Read the full abstract

All authors Yilai Liu, Wenqi Pei, Zhifeng Lu, Shiyuan Zhang, Zongxi Liu, Hongyang Du

Solar and Stellar Astrophysics

Formation Dynamics of a Rotating Solar Spicule-like Event and Associated Energy-Mass Transport

Liufan Gong, Lei Ni, Guanchong Cheng et al. (+3 more)

We performed three-dimensional radiative magnetohydrodynamic(3D RMHD) simulations in an upward-directed (positive-polarity) open magnetic field with an initial background strength of 10 G. We identified vortex-associated spicule-like events and carried out a detailed analysis of one representative event. Prior to the formation of this spicule-like event, a photospheric vortex appears and twists the magnetic field lines, then develops upward into higher atmospheric layers, forming a magnetic tornado system. This event is generated during this process. At the periphery of this magnetic tornado, twisted magnetic field lines with negative vertical components reconnect with both external positive-polarity background fields and central positive-polarity fields. The magnetic reconnection sites rise and rotate together with the developed magnetic tornado, and many jet-like substructures in this event are closely associated with plasma upward ejections from these reconnection regions. A vertical force analysis shows that, for this event, plasma uplift in the photosphere and lower chromosphere is dominated by the pressure-gradient force, whereas in the upper chromosphere, the Lorentz force gradually becomes dominant. At a height of about 4 Mm above the solar surface, this event produces a mean upward mass flux density on the order of $10^{-7}~{\rm kg~m^{-2}~s^{-1}}$ and a mean upward Poynting flux density of approximately $525~{\rm W~m^{-2}}$. The wave-mode decomposition results show that the energy transport associated with the rotational spicule-like event is dominated by the Alfvén mode. Alfvén waves, fast-mode waves, and slow-mode waves carry energy into the lower corona in an approximate ratio of $10:7:3$.
Read the full abstract

All authors Liufan Gong, Lei Ni, Guanchong Cheng, Jiajia Liu, Quan Xie, Robert Erdélyi