arXiv subjects

Lingfeng Li

Publications and source records attributed to Lingfeng Li.

At least 19 recordsLinked to original sources

arXiv2026

Inelastic from the Other Side: Xenon Excitation Signals in Light of the LZ High-Recoil Event

The LUX-ZEPLIN (LZ) experiment recently reported a single event consistent with a nuclear recoil of $E_{\rm NR}\sim248\pm23\,(\mathrm{stat})\pm23\,(\mathrm{syst})$~keV in an extended-energy search. We study the accompanying inelastic-xenon channel, $\chi+\Xe\to\chi+\Xe^{*}$, which provides a complementary test of DM interpretations of such high-recoil events. Using the non-relativistic effective-field-theory (NREFT) framework with shell-model nuclear-transition inputs, we compute the inelastic-xenon rates for representative scenarios including inelastic dark matter and accelerated DM populations. We also simulate their S1--S2 responses, with detector modeling validated against public LZ data. The nuclear de-excitation adds an electromagnetic component shifting the signal toward the electronic-recoil band and, in some cases, toward clustered backgrounds from isotopes. We therefore construct a schematic, physics-motivated background model and perform a simplified statistical analysis. For many benchmarks, the inelastic-xenon rate is $\mathcal{O}(0.1)$ of the elastic rate or below, suggesting that substantially larger exposures are required before this companion channel becomes observable. Our results illustrate the importance of isotope-induced background structures in the extended-energy region and extend the phenomenology of inelastic-xenon signatures to a wider range of DM scenarios.

hep-ph↗

arXiv2026

A Unified Variational Framework for Deep Weakly Supervised Image Segmentation

We propose a unified variational framework for image segmentation under sparse pixel-level supervision. Our method is based on a simplex-constrained Potts model with a smooth perimeter regularizer, yielding a convex, smooth energy functional that can be used as a training loss in weakly supervised deep learning paradigms or optimized efficiently using iterative methods. Sparse labels are incorporated into the data fidelity term by constructing a fuzzy membership function via a function extension problem in a Reproducing Kernel Hilbert Space (RKHS), which can effectively capture inhomogeneous intensity statistics. The derived discrete loss for training standard networks demonstrates robustness and consistent improvements over non-training and partial cross-entropy (PCE) baselines in experiments, achieving comparable performance without requiring ground-truth segmentation images.

cs.CV↗

arXiv2026

MoWorld: A Flash World Model

The future of World Models depends not only on scaling model capability, but also on scaling practicality and inference efficiency. High-frame-rate inference enables responsive perception, planning, and control in real-world autonomous systems. To this end, we present MoWorld, a cost-effective yet high-performance Flash World Model with an end-to-end framework spanning data generation, pre-training, distillation, and efficient inference, enabling up to 50 FPS real-time interaction with cinematic visual quality without the need of high-end GPUs. To enable large-scale real-world deployment, MoWorld jointly optimizes model capability and cost throughout the entire development pipeline. Specifically, unlike existing approaches that primarily rely on large-scale video corpora, MoWorld is built upon a scalable 3D-native data engine accumulated from our large-scale 3D vision and generative modeling pipeline, enabling the efficient construction of geometrically consistent training data across diverse real-world and synthetic environments. Based on this foundation, a curriculum cross-frame pre-training strategy for stable and scalable World Model learning, an efficient denoising-step distillation algorithm to reduce diffusion training cost, and a mixed-precision parallel inference framework for low-cost real-time deployment. MoWorld is the first real-time interactive World Model built on the Neural Processing Unit (NPU) and can achieves up to 50 FPS in such the devices, enabling practical and efficient deployment at scale. Comprehensive evaluations demonstrate that MoWorld achieves leading performance; notably, its average inference cost is only 30\%-50\% of that of existing World Models, providing a practical foundation for large-scale real-world applications of World Models. We also demonstrate diverse applications of MoWorld.

cs.CV↗

arXiv2026

Probing a 146 GeV cLFV scalar using the LHC and low-energy experiments

The CMS Collaboration reported a local excess at $146~\mathrm{GeV}$ in the search for the lepton-flavor-violating decay of the Higgs boson and additional Higgs bosons in the $e\mu$ final state at $\sqrt{s}= 13~\mathrm{TeV}$. If confirmed, this would constitute a major piece of evidence of charged lepton flavor violation (cLFV). We investigate the compatibility of the claimed signal with the full suite of existing low-energy cLFV constraints in a bottom-up effective description: a single real scalar of mass $146~\mathrm{GeV}$ coupled to gluons and to all charged-lepton bilinears, with seven free parameters that simultaneously control the LHC production cross section, every di-lepton decay channel, and every low-energy cLFV observable. A Bayesian MCMC analysis against $\mu-e$ conversion, muonium-antimuonium oscillation, three-lepton and radiative LFV decays, semileptonic $\tau$ LFV decays, and LHC di-lepton searches yields a preferred mode with peaked value $Y_{e\mu} \sim 10^{-4.09}$, already cut into by the current $\mu-e$ conversion limits. The projected sensitivities of Mu2e, COMET, Mu3e, MACE, MEG~II, Belle~II, STCF, and the HL-LHC directly probe the region of coupling space selected by the CMS excess, so the complementarity between high-energy and low-energy cLFV probes will either corroborate or decisively exclude the scalar interpretation of the anomaly within the next decade.

hep-ph↗

arXiv2026

CMB Constraints on Pre-Inflationary Axion Dark Matter Isocurvature

Although measurements of the Cosmic Microwave Background (CMB) are consistent with a nearly scale-invariant primordial spectrum of adiabatic perturbations, in which the energy densities of different components (radiation, baryons, and dark matter) fluctuate proportionally, there could also exist isocurvature perturbations, in which density fluctuations of the individual components differ from the adiabatic mode. Cold dark matter isocurvature (CDI) perturbations with a variety of spectral tilts generated in pre-inflationary axion models provide one such example. In this article, we present the most updated constraints on these axion CDI perturbations using the latest CMB anisotropy measurements from Planck, the Atacama Cosmology Telescope (ACT), and the South Pole Telescope (SPT). We study both fixed spectral indices with values ranging from red- to blue-tilted spectra as well as the case with a free index. We find that the constraint on the spectral index gets moderately improved with the combined datasets compared to Planck alone, while the bounds on the isocurvature amplitudes for the fixed spectral indices we consider do not get tighter. We also discuss the theoretical implications of our constraints, in particular for models giving rise to blue-tilted spectra.

astro-ph.CO↗

arXiv2026

Solving Convolution-type Integral Equations using Preconditioned Neural Operators

Convolution-type integral equations arise from various fields, \textit{e.g.}, finite impulse response filters in signal processing and deblurring problems in image processing. When solving these equations, conventional numerical methods, like the multigrid method, can only efficiently solve the low-frequency components in the error, but not the high-frequency components. In this paper, we apply neural operators to address this issue. By adopting a preconditioning approach, we propose a novel training strategy that trains neural operators to solve the high-frequency components efficiently. Then, we combine the neural operators with some classical iterative solvers, like the weighted Jacobi method, to obtain an efficient hybrid iterative algorithm for the integral equations. We analyze the generalization error of our training strategy and the convergence of the hybrid iterative algorithm. We test our algorithms on large-scale and ill-conditioned linear systems discretized from one- and two-dimensional convolution-type integral equations. Our proposed algorithm significantly outperforms the multigrid method and the preconditioned conjugate gradient method in both iteration numbers and computational time.

math.NA↗

arXiv2026

Beyond Autoregressive RTG: Conditioning via Injection Outside Sequential Modeling in Decision Transformer

Decision Transformer (DT) formulates offline reinforcement learning as autoregressive sequence modeling, achieving promising results by predicting actions from a sequence of Return-to-Go (RTG), state, and action tokens. However, RTG is a scalar that summarizes future rewards, containing far less information than typical state or action vectors, yet it consumes the same computational budget per token. Worse, the self-attention cost of Transformers grows quadratically with sequence length, so including RTG as a separate token adds unnecessary overhead. We propose SlimDT, which removes RTG from the autoregressive sequence. Instead, we inject RTG information into the state representations before the sequential modeling step, allowing the Transformer to process only a compact (state, action) sequence. This reduces the sequence length by one-third, directly improving inference efficiency. On the D4RL benchmark, SlimDT surpasses standard DT across various tasks and achieves performance comparable to existing state-of-the-art methods. Decoupling a sparse conditioning signal from an information-rich sequence thus yields both computational gains and higher task performance.

cs.LG↗

arXiv2026

In-context modeling as a retrain-free paradigm for foundation models in computational science

Building models that generalize across physical systems without retraining remains a central challenge in computational science. Here we introduce In-Context Modeling (ICM), a retrain-free paradigm that infers physical relationships directly from observational fields. Rather than encoding system-specific behavior in fixed parameters, ICM assimilates measurements as physical context and performs inference through a single forward pass. Trained in a physics-informed, label-free manner using governing equations, a single model generalizes across unseen materials, geometries, and loading conditions. Demonstrated on hyperelasticity, ICM integrates with finite-element simulations and is validated using experimental full-field measurements. Moreover, performance improves with increasing data diversity and computational budget, exhibiting favorable scaling behavior analogous to foundation models. By recasting physical modeling as in-context inference, this work establishes a transferable paradigm for retrain-free scientific learning and a foundation for scalable modeling across computational science.

cs.CE↗

arXiv2026

Dark Matter on a Slide

We present a scenario for GeV-scale thermal dark matter that can only be tested with accelerator experiments. Dark matter is composed of dark pions arising from a confining strong interaction in the dark sector. The thermal relic density is obtained through the interplay of up-scatterings of dark pions to heavier dark mesons (the dark counterparts of the kaons and $\eta$), and decays of the unstable dark $\eta$ to Standard Model particles. This mechanism is analogous to a playground slide, where one climbs up first and then slides down with a release of energy. We illustrate the scenario with a minimal model based on the SU(3)/SO(3) coset, where dark matter is stabilized by a U(1) flavor symmetry. The correct relic density is obtained with dark meson mass splittings of 10% to 50% and a dark-$\eta$ lifetime shorter than $10^3\,\mathrm{m}/c$. Direct and indirect dark matter searches are mostly ineffective, as a consequence of the charge conjugation symmetry of the stabilizing U(1). The most striking signals arise at the LHC, from the production of dark showers containing long-lived dark $\eta$'s that decay to visible final states. These signatures crucially depend on the portal interaction connecting the dark sector to the Standard Model. We show that several well-known portals can complete the scenario above the weak scale, and outline the expected signals in each case.

hep-ph↗

arXiv2026

Exotic Higgs Decays at a Muon Collider

We study the sensitivity of a future muon collider to exotic Higgs decays in a minimal scenario of Standard Model (SM) augmented with a light singlet scalar $S$. We consider the decay $h\to SS$ and $S$'s subsequently decay back to SM. In particular, we focus on final states with four bottom quarks ($4b$), or two bottom quarks and two muons ($2b2\mu$). Analyses are performed for two muon collider benchmark configurations: center-of-mass collision energy $\sqrt{s}=3~\mathrm{TeV}$ with $1~\mathrm{ab}^{-1}$ data and $\sqrt{s}=10~\mathrm{TeV}$ with $10~\mathrm{ab}^{-1}$ data. Machine-learning techniques are applied to suppress backgrounds and mitigate jet-combinatorics effects in both channels. We find that the $4b$ mode could be sensitive to the branching ratio, BR$(h \to SS \to 4b)$, of ${\cal O}(10^{-2})$ at 3 TeV and ${\cal O}(10^{-3})$ at 10 TeV, significantly improving upon high-luminosity LHC projections. In the Higgs-portal model with $S$ coupling to SM only through mixing with the Higgs, the sensitivities to BR$(h \to SS)$ remain at the same level given ${\cal O}(1)$ branching fraction of $S$ decaying into $b$-quarks. The $2b2\mu$ mode benefits from a clean dimuon resonance and can probe BR$(h\to SS\to 2b2\mu)$ down to $10^{-5}$ level at a 10 TeV muon collider. But the sensitivity to BR$(h \to SS)$ will be significantly reduced due to the small branching fraction of $S$ decaying into muons in the Higgs portal model.

hep-ph↗

arXiv2026

System Design for Maintaining Internal State Consistency in Long-Horizon Robotic Tabletop Games

Long-horizon tabletop games pose a distinct systems challenge for robotics: small perceptual or execution errors can invalidate accumulated task state, propagate across decision-making modules, and ultimately derail interaction. This paper studies how to maintain internal state consistency in turn-based, multi-human robotic tabletop games through deliberate system design rather than isolated component improvement. Using Mahjong as a representative long-horizon setting, we present an integrated architecture that explicitly maintains perceptual, execution, and interaction state, partitions high-level semantic reasoning from time-critical perception and control, and incorporates verified action primitives with tactile-triggered recovery to prevent premature state corruption. We further introduce interaction-level monitoring mechanisms to detect turn violations and hidden-information breaches that threaten execution assumptions. Beyond demonstrating complete-game operation, we provide an empirical characterization of failure modes, recovery effectiveness, cross-module error propagation, and hardware-algorithm trade-offs observed during deployment. Our results show that explicit partitioning, monitored state transitions, and recovery mechanisms are critical for sustaining executable consistency over extended play, whereas monolithic or unverified pipelines lead to measurable degradation in end-to-end reliability. The proposed system serves as an empirical platform for studying system-level design principles in long-horizon, turn-based interaction.

cs.RO↗

arXiv2026

A level-wise training scheme for learning neural multigrid smoothers with application to integral equations

Convolution-type integral equations commonly occur in signal processing and image processing. Discretizing these equations yields large and ill-conditioned linear systems. While the classic multigrid method is effective for solving linear systems derived from partial differential equations (PDE) problems, it fails to solve integral equations because its smoothers, which are implemented as conventional relaxation methods, are ineffective in reducing high-frequency components in the errors. We propose a novel neural multigrid scheme where learned neural operators replace classical smoothers. Unlike classical smoothers, these operators are trained offline. Once trained, the neural smoothers generalize to new right-hand-side vectors without retraining, making it an efficient solver. We design level-wise loss functions incorporating spectral filtering to emulate the multigrid frequency decomposition principle, ensuring each operator focuses on solving distinct high-frequency spectral bands. Although we focus on integral equations, the framework is generalizable to all kinds of problems, including PDE problems. Our experiments demonstrate superior efficiency over classical solvers and robust convergence across varying problem sizes and regularization weights.

cs.LG↗

arXiv2026

Primordial features as probes of baryogenesis from supersymmetric flat directions

The Affleck-Dine mechanism is a leading baryogenesis scenario in which scalar condensates form coherently during inflation along supersymmetric flat directions that are lifted by supersymmetry-breaking effects. We update the viable parameter space for baryogenesis using recent Cosmic Microwave Background constraints on baryon-density isocurvature perturbations, taking the quantum fluctuations of the scalar condensate generated during inflation as initial conditions. We then show that primordial features arising from the inflaton sector can serve as a unique probe of baryogenesis models, whose mechanisms are otherwise difficult to access directly due to their high energy scales. These primordial features leave correlated imprints, such as sharp feature signals and clock signals, on both the curvature and baryon-density isocurvature perturbations, providing direct evidence for the existence of both light and heavy modes involved in the Affleck-Dine mechanism.

hep-ph↗

arXiv2026

Pareto-guided Pipeline for Distilling Featherweight AI Agents in Mobile MOBA Games

Recent advances in game AI have demonstrated the feasibility of training agents that surpass top-tier human professionals in complex environments such as Honor of Kings (HoK), a leading mobile multiplayer online battle arena (MOBA) game. However, deploying such powerful agents on mobile devices remains a major challenge. On one hand, the intricate multi-modal state representation and hierarchical action space of HoK demand large, sophisticated policy networks that are inherently difficult to compress into lightweight forms. On the other hand, production deployment requires high-frequency inference under strict energy and latency constraints on mobile platform. To the best of our knowledge, bridging large-scale game AI and practical on-device deployment has not been systematically studied. In this work, we propose a Pareto optimality guided pipeline and design a high-efficiency student architecture search space tailored for mobile execution, enabling systematic exploration of the trade-off between performance and efficiency. Experimental results demonstrate that the distilled model achieves remarkable efficiency, including an $12.4\times$ faster inference speed (under 0.5ms per frame) and a $15.6\times$ improvement in energy efficiency (under 0.5mAh per game), while retaining a 40.32% win rate against the original teacher model.

cs.LG↗

arXiv2026

Decoupling Return-to-Go for Efficient Decision Transformer

The Decision Transformer (DT) has established a powerful sequence modeling approach to offline reinforcement learning. It conditions its action predictions on Return-to-Go (RTG), using it both to distinguish trajectory quality during training and to guide action generation at inference. In this work, we identify a critical redundancy in this design: feeding the entire sequence of RTGs into the Transformer is theoretically unnecessary, as only the most recent RTG affects action prediction. We show that this redundancy can impair DT's performance through experiments. To resolve this, we propose the Decoupled DT (DDT). DDT simplifies the architecture by processing only observation and action sequences through the Transformer, using the latest RTG to guide the action prediction. This streamlined approach not only improves performance but also reduces computational cost. Our experiments show that DDT significantly outperforms DT and establishes competitive performance against state-of-the-art DT variants across multiple offline RL tasks.

cs.AI↗

arXiv2026

BotzoneBench: Scalable LLM Evaluation via Graded AI Anchors

Large Language Models (LLMs) are increasingly deployed in interactive environments requiring strategic decision-making, yet systematic evaluation of these capabilities remains challenging. Existing benchmarks for LLMs primarily assess static reasoning through isolated tasks and fail to capture dynamic strategic abilities. Recent game-based evaluations employ LLM-vs-LLM tournaments that produce relative rankings dependent on transient model pools, incurring quadratic computational costs and lacking stable performance anchors for longitudinal tracking. The central challenge is establishing a scalable evaluation framework that measures LLM strategic reasoning against consistent, interpretable standards rather than volatile peer models. Here we show that anchoring LLM evaluation to fixed hierarchies of skill-calibrated game Artificial Intelligence (AI) enables linear-time absolute skill measurement with stable cross-temporal interpretability. Built on the Botzone platform's established competitive infrastructure, our BotzoneBench evaluates LLMs across eight diverse games spanning deterministic perfect-information board games to stochastic imperfect-information card games. Through systematic assessment of 177,047 state-action pairs from five flagship models, we reveal significant performance disparities and identify distinct strategic behaviors, with top-performing models achieving proficiency comparable to mid-to-high-tier specialized game AI in multiple domains. This anchored evaluation paradigm generalizes beyond games to any domain with well-defined skill hierarchies, establishing a scalable and reusable framework for assessing interactive AI capabilities.

cs.AI↗

arXiv2026

Adapting Rules of Official International Mahjong for Online Players

As one of the worldwide spread traditional game, Official International Mahjong can be played and promoted online through remote devices instead of requiring face-to-face interaction. However, online players have fragmented playtime and unfixed combination of opponents in contrary to offline players who have fixed opponents for multiple rounds of play. Therefore, the rules designed for offline players need to be modified to ensure the fairness of online single-round play. Specifically, We employ a world champion AI to engage in self-play competitions and conduct statistical data analysis. Our study reveals the first-mover advantage and issues in the subgoal scoring settings. Based on our findings, we propose rule adaptations to make the game more suitable for the online environment, such as introducing compensatory points for the first-mover advantage and refining the scores of subgoals for different tile patterns. Compared with the traditional method of rotating positions over multiple rounds to balance first-mover advantage, our compensatory points mechanism in each round is more convenient for online players. Furthermore, we implement the revised Mahjong game online, which is open for online players. This work is an initial attempt to use data from AI systems to evaluate Official Internatinoal Mahjong's game balance and develop a revised version of the traditional game better adapted for online players.

cs.AI↗

arXiv2025

Prospect for measurement of CP-violating parameters of $B_s^0 \to \phi\gamma$ at the Tera Z factory

$b \to s\gamma$ transition is a critical flavor-changing neutral current (FCNC) process that could be used to probe CP violation (CPV) and new physics (NP). We quantify the anticipated precision for measuring $B_s^0 \to \phi\gamma$ at the CEPC Z pole operation, showing that the relative statistical uncertainty could be as low as 0.16\%, improved by approximately two orders of magnitude compared to existing measurements. Additionally, we perform a time-dependent analysis of the $B_s^0 \to \phi\gamma$ decay, accounting for $B_s^0/\bar{B}_s^0$ mixing extract the mixing-induced and CP-violating parameters $\boldsymbol{\mathcal{A}_{\phi\gamma}^\Delta}$, $\boldsymbol{C_{\phi\gamma}}$ and $\boldsymbol{S_{\phi\gamma}}$. Using central value from LHCb measurement as input, we evaluate the anticipated accuracy of measurements of these parameters. The projected statistical uncertainties are $\sigma_{A_{\phi\gamma}^{\Delta}{}^{\text{stat}}} = 0.021$, $\sigma_C^{\text{stat}} = 0.0092$ and $\sigma_S^{\text{stat}} = 0.0096$, and the systematic uncertainties are $\sigma_{A_{\phi\gamma}^{\Delta}{}^{\text{syst}}} = 0.035$, $\sigma_C^{\text{syst}} = 0.0027$ and $\sigma_S^{\text{syst}} = 0.0064$. Furthermore, the 1$\sigma$ sensitivity boundaries for NP in this study are found to be $\mathcal{A}_{\phi\gamma}^\Delta < -0.05$ or $\mathcal{A}_{\phi\gamma}^\Delta > 0.15$, $\mathcal{C}_{\phi\gamma} < -0.02$ or $\mathcal{C}_{\phi\gamma} > 0.04$, and $\mathcal{S}_{\phi\gamma} < -0.04$ or $\mathcal{S}_{\phi\gamma} > 0.04$. We also conduct a relevant detector optimization study by establishing the correlation between the anticipated precision and the intrinsic resolution of the ECAL, as well as the performance of the PID system.

hep-ex↗