Searcharxiv⌕ Search

arXiv subjects

Siyu Jiang

Publications and source records attributed to Siyu Jiang.

15 recordsLinked to original sources
arXiv2026

Multimessenger search strategy for composite dark matter with white dwarf data and gravitational wave detectors

The nature of dark matter (DM) remains one of the most challenges in modern physics. Composite or macroscopic DM present a compelling alternative to conventional particle DM, yet their terrestrial search is notoriously challenging due to low number density. This Letter presents a unified, multimessenger search strategy for composite DM, dramatically improving existing constraints and proposing a new detection method. We first perform a model independent update of astrophysical constraints from compact objects, utilizing systematic calculations and additional white dwarf data related to ignition and subsequent supernovae. Crucially, for the first time, we explore novel signals of composite DM in future gravitational wave detectors like LISA, TianQin, and Taiji, performing detailed signal to noise ratio and Fisher matrix analyses. We demonstrate that these detectors will possess the requisite sensitivity to probe untouched regions of the DM parameter space. Our results underscore the unique power of the multimessenger paradigm spanning stellar astrophysics and gravitational wave astronomy to explore this distinct and challenging frontier of DM physics. Our analyses extend broadly to a wide range of macroscopic or composite DM scenarios.

astro-ph.HE↗
arXiv2026

Freeze-in gravitational waves and dark matter in warm inflation

Recent study [1] has suggested that warm inflation may be realized with a minimal extension of the Standard Model by a single scalar inflaton field with an axion-like coupling to gluons. Motivated by this framework, we investigate the gravitational wave spectrum and graviton-portal dark matter production through the freeze-in process generated during warm inflation scenarios. We perform a comparative analysis for different dissipation terms, focusing on their distinct gravitational wave signatures in the high-frequency regime. Our findings reveal qualitative and quantitative differences in the spectral behavior, offering a preliminary pathway for exploring inflationary and dark matter models through high-frequency gravitational wave signals.

hep-ph↗
arXiv2026

KVSmooth: Mitigating Hallucination in Multi-modal Large Language Models through Key-Value Smoothing

Despite the significant progress of Multimodal Large Language Models (MLLMs) across diverse tasks, hallucination -- corresponding to the generation of visually inconsistent objects, attributes, or relations -- remains a major obstacle to their reliable deployment. Unlike pure language models, MLLMs must ground their generation process in visual inputs. However, existing models often suffer from semantic drift during decoding, causing outputs to diverge from visual facts as the sequence length increases. To address this issue, we propose KVSmooth, a training-free and plug-and-play method that mitigates hallucination by performing attention-entropy-guided adaptive smoothing on hidden states. Specifically, KVSmooth applies an exponential moving average (EMA) to both keys and values in the KV-Cache, while dynamically quantifying the sink degree of each token through the entropy of its attention distribution to adaptively adjust the smoothing strength. Unlike computationally expensive retraining or contrastive decoding methods, KVSmooth operates efficiently during inference without additional training or model modification. Extensive experiments demonstrate that KVSmooth significantly reduces hallucination ($\mathit{CHAIR}_{S}$ from $41.8 \rightarrow 18.2$) while improving overall performance ($F_1$ score from $77.5 \rightarrow 79.2$), achieving higher precision and recall simultaneously. In contrast, prior methods often improve one at the expense of the other, validating the effectiveness and generality of our approach.

cs.CV↗
arXiv2026

Knob: A Physics-Inspired Gating Interface for Interpretable and Controllable Neural Dynamics

Existing neural network calibration methods often treat calibration as a static, post-hoc optimization task. However, this neglects the dynamic and temporal nature of real-world inference. Moreover, existing methods do not provide an intuitive interface enabling human operators to dynamically adjust model behavior under shifting conditions. In this work, we propose Knob, a framework that connects deep learning with classical control theory by mapping neural gating dynamics to a second-order mechanical system. By establishing correspondences between physical parameters -- damping ratio ($ζ$) and natural frequency ($ω_n$) -- and neural gating, we create a tunable "safety valve". The core mechanism employs a logit-level convex fusion, functioning as an input-adaptive temperature scaling. It tends to reduce model confidence particularly when model branches produce conflicting predictions. Furthermore, by imposing second-order dynamics (Knob-ODE), we enable a \textit{dual-mode} inference: standard i.i.d. processing for static tasks, and state-preserving processing for continuous streams. Our framework allows operators to tune "stability" and "sensitivity" through familiar physical analogues. This paper presents an exploratory architectural interface; we focus on demonstrating the concept and validating its control-theoretic properties rather than claiming state-of-the-art calibration performance. Experiments on CIFAR-10-C validate the calibration mechanism and demonstrate that, in Continuous Mode, the gate responses are consistent with standard second-order control signatures (step settling and low-pass attenuation), paving the way for predictable human-in-the-loop tuning.

cs.AI↗
arXiv2026

A New Source of Phase Transition Gravitational Waves: Heavy Particle Braking Across Bubble Walls

Motivated by the new heavy dark matter production mechanism from cosmic phase transition, we propose a novel mechanism for the generation of microscopic gravitational waves (GWs) during cosmological first-order phase transitions arising from the braking of heavy particles as they traverse bubble walls. Unlike the well-known sources such as bubble collisions, sound waves, or turbulence in the plasma, this mechanism originates from the direct interaction between massive particles and the expanding bubble wall. We use quantum field theory to rigorously compute the gravitational radiation. The resulting GW spectrum exhibits distinctive features: The peak frequency is tightly correlated with the bubble wall velocity, while the peak amplitude scales as the fourth power of the heavy particle mass. These unique dependencies offer a new observational handle on particle physics beyond the Standard Model. We illustrate this mechanism within a specific model framework and demonstrate its viability. Our findings enrich the landscape of phase transition GW sources and open new avenues for more directly probing heavy particle dynamics and new physics models in the early universe.

hep-ph↗
arXiv2025

Context Consistency Learning via Sentence Removal for Semi-Supervised Video Paragraph Grounding

Semi-Supervised Video Paragraph Grounding (SSVPG) aims to localize multiple sentences in a paragraph from an untrimmed video with limited temporal annotations. Existing methods focus on teacher-student consistency learning and video-level contrastive loss, but they overlook the importance of perturbing query contexts to generate strong supervisory signals. In this work, we propose a novel Context Consistency Learning (CCL) framework that unifies the paradigms of consistency regularization and pseudo-labeling to enhance semi-supervised learning. Specifically, we first conduct teacher-student learning where the student model takes as inputs strongly-augmented samples with sentences removed and is enforced to learn from the adequately strong supervisory signals from the teacher model. Afterward, we conduct model retraining based on the generated pseudo labels, where the mutual agreement between the original and augmented views' predictions is utilized as the label confidence. Extensive experiments show that CCL outperforms existing methods by a large margin.

cs.CV↗
arXiv2025

Pseudo-Goldstone Dark Matter from Primordial Black Holes: Gravitational Wave Signatures and Implications for KM3-230213A Event at KM3NeT

In many well-motivated new physics models, the pseudo-Nambu-Goldstone boson (pNGB) from U(1) symmetry breaking emerges as a promising dark matter candidate. Its coupling, suppressed by the symmetry breaking scale, prevents thermal equilibrium in the early Universe for high scale symmetry breaking. Thus, pNGB dark matter is predominantly produced via non-thermal mechanisms, such as the freeze-in process through a new portal coupling. In this work, we explore a novel mechanism for the production of pNGB dark matter even with feeble Higgs portal coupling-arising from Hawking radiation or superradiance of primordial black holes. We systematically investigate the production of light and heavy pNGB dark matter, both for Schwarzschild and Kerr black holes. We also discuss its potential gravitational wave signatures from domain wall collapse, density perturbations, and Hawking radiation. If the ultraviolet (UV) model is considered, the recent $\mathcal{O}$(100) PeV neutrino event KM3-230213A at KM3NeT can be naturally explained.

hep-ph↗
arXiv2024

Gauged Q-ball dark matter through a cosmological first-order phase transition

As a new type of dynamical dark matter mechanism, we discuss the stability of the gauged Q-ball dark matter and its production mechanism through a cosmological first-order phase transition. This work delves into the study of gauged Q-ball dark matter generated during the cosmic phase transition. We demonstrate detailed discussions on the stability of gauged Q-balls to rigorously constrain their charge and mass ranges. Additionally, employing analytic approximations and the mapping method, we provide qualitative insights into gauged Q-balls. We establish an upper limit on the gauge coupling constant and give the relic density of stable gauged Q-ball dark matter formed during a first-order phase transition. Furthermore, we discuss potential observational signatures or constraints of gauged Q-ball dark matter, including astronomical observations and gravitational wave signals.

hep-ph↗
arXiv2024

NTFormer: A Composite Node Tokenized Graph Transformer for Node Classification

Recently, the emerging graph Transformers have made significant advancements for node classification on graphs. In most graph Transformers, a crucial step involves transforming the input graph into token sequences as the model input, enabling Transformer to effectively learn the node representations. However, we observe that existing methods only express partial graph information of nodes through single-type token generation. Consequently, they require tailored strategies to encode additional graph-specific features into the Transformer to ensure the quality of node representation learning, limiting the model flexibility to handle diverse graphs. To this end, we propose a new graph Transformer called NTFormer to address this issue. NTFormer introduces a novel token generator called Node2Par, which constructs various token sequences using different token elements for each node. This flexibility allows Node2Par to generate valuable token sequences from different perspectives, ensuring comprehensive expression of rich graph features. Benefiting from the merits of Node2Par, NTFormer only leverages a Transformer-based backbone without graph-specific modifications to learn node representations, eliminating the need for graph-specific modifications. Extensive experiments conducted on various benchmark datasets containing homophily and heterophily graphs with different scales demonstrate the superiority of NTFormer over representative graph Transformers and graph neural networks for node classification.

cs.LG↗
arXiv2024

Implication of nano-Hertz stochastic gravitational wave on dynamical dark matter through a dark first-order phase transition

For the first time, the expected stochastic gravitational wave background is probably discovered after observing the Hellings Downs correlation curve by several pulsar timing array (PTA) collaborations around the globe including NANOGrav, European PTA, Parkes PTA, and Chinese PTA. These new observations can help to explore or constrain the dark matter (DM) formation mechanisms in the early Universe. We study the implication of those results on the dynamical DM formation mechanisms through a dark first-order phase transition in the early Universe. Both the Q-ball DM and super-cool DM are investigated in the strong super-cooling dark phase transition scenario which may give an interpretation of the observed stochastic gravitational wave background.

hep-ph↗
arXiv2023

Hydrodynamic effects on the filtered dark matter produced by a first-order phase transition

Motivated by current status of dark matter (DM) search, a new type of DM production mechanism is proposed based on thedynamical process of a strong first-order phase transition in the early universe, namely, the filtered DM mechanism. We study the hydrodynamic effects on the DM relic density. By detailed calculations, we demonstrate that the hydrodynamic modes with the corresponding hydrodynamic heating effects play essential roles in determining the DM relic density. The corresponding phase transition gravitational wave could help to probe this new mechanism.

hep-ph↗
arXiv2021

Pseudo-Dirac Sterile Neutrino Dark Matter

Sterile neutrino is a promising dark matter (DM) candidate. However the parameter space of this scenario has almost been ruled out by the X-ray observation results whenever the sterile neutrino is solely produced by the Dodelson-Widrow (DW) mechanism in the early Universe. In this letter we propose an extension to the minimal sterile neutrino DM model by introducing the pseudo-Dirac sterile neutrino, which implies the existence of two nearly degenerate Majorana states $\hat N_{1,2}^{}$, and a permutation symmetry. The heavy state $\hat N_1$ is produced via the DW mechanism, and the light state $\hat N_2$, which serves as the DM, is produced from the decay of $\hat N_1$ in the early Universe. The X-ray constraint is avoided by the permutation symmetry, which forbidden the two-body decay of the DM into active neutrinos and photon. A promising signal of this scenario is the effective number of neutrino species, which will be precisely measured in future experiments, such as CMB stage IV. We further study the impact of this model on the cosmological parameters. The Markov Chain Monte Carlo analysis for the Planck + BAO+R19 data gives $H_0=69.2_{-0.59}^{+0.58}$, which may relieve the Hubble tension.

hep-ph↗
arXiv2020

Neutrino oscillation in dark matter with $L_μ-L_τ$

In this paper, we study the phenomenology of a Dirac dark matter in the $L_μ-L_τ$ model and investigate the neutrino oscillation in the dark halo. Since dark matter couples to the muon neutrino and the tau neutrino with opposite sign couplings, it contributes effective potentials, $\pm A_χ$, to the evolution equation of the neutrino flavor transition amplitude, which can be significant for high energy neutrino oscillations in a dense dark matter environment. We discuss neutrino masses, lepton mixing angles, Dirac CP phase, and neutrino oscillation probabilities in the dark halo using full numerical calculations. Results show that neutrinos can endure very different matter effects. When the potential $A_χ$ becomes ultra-large, three neutrino flavors decouple from each other.

hep-ph↗
arXiv2018

Battery-like Supercapacitors from Vertically Aligned Carbon Nanofibers Coated Diamond: Design and Demonstrator

Battery-like supercapacitors feature high power and energy densities as well as long-term capacitance retention. The utilized capacitor electrodes are thus better to have large surface areas, high conductivity, high stability, and importantly be of binder free. Herein, vertically aligned carbon nanofibers (CNFs) coated boron-doped diamonds (BDD) are employed as the capacitor electrodes to construct battery-like supercapacitors. Grown via a thermal chemical vapor deposition technique, these CNFs/BDD hybrid films are binder free and own porous structures, resulting in large surface areas. Meanwhile, the containment of graphene layers and copper metal catalysts inside CNFs/BDD leads to their high conductivity. Electric double layer capacitors (EDLCs) and pseudocapacitors (PCs) are then constructed in the inert electrolyte (1.0 M H2SO4 solution) and in the redox-active electrolyte (1.0 M Na2SO4 + 0.05 M Fe(CN)63-/4-), respectively. For assembled two-electrode symmetrical supercapacitor devices, the capacitances of EDLC and PC devices reach 30 and 48 mF cm-2 at 10 mV s-1, respectively. They remain constant even after 10 000 cycles. The power densities are 27.3 kW kg-1 and 25.3 kW kg-1 for EDLC and PC devices, together with their energy densities of 22.9 Wh kg-1 and 44.1 Wh kg-1, respectively. The performance of formed EDLC and PC devices is comparable to market-available batteries. Therefore, the vertically aligned CNFs/BDD hybrid film is a suitable capacitor electrode material to construct high-performance battery-like and industry-orientated supercapacitors for flexible power devices.

physics.app-ph↗