BZPEERPapers, people, evidence and research paths
Menu

Discover

SearchFind papers, people, topics, institutions and opportunities from one place.GraphSee how papers, authors, topics and evidence connect.

Research tools

HomeContinue from the papers, people and topics that matter now.CollectionsTurn saved papers into reading lists, evidence maps and shareable context.ResearchReconstruct the path that led you from search to paper to question.CollaborationMove from a paper, topic or expert request into a focused thread.PublishCreate full-text research articles with metadata, formulas and versions.

Network

ExpertsFind specialists whose work and reviews explain why they can help.CommunitiesJoin research groups organized around papers, topics and shared evidence.

Opportunities

ListingsBrowse roles, calls and collaborations connected to real research context.

Account

Log inReturn to your saved papers, graph views and active threads.Create accountStart saving papers, building a research trail and joining teams.
Log inCreate account

Source author record

Hyungjin Kim

Hyungjin Kim appears in the imported research catalog. Authorship, coauthor and topic links are available while profile ownership is still unclaimed.

ResearcherUnclaimed source record
hep-phastro-ph.COphysics.atom-phhep-thMachine LearningArtificial Intelligenceastro-ph.GAphysics.app-phphysics.optics

Catalog footprint

What is connected

13works
9topics
4close collaborators

Actions

Connect this record

Log in to claim
Open graphBrowse works

Research graph

See the researcher in context

Open full explorer

Inspect adjacent papers, topics, institutions and collaborators without losing the researcher page.

Building this map preview

BZPEER is loading the nearby papers, people, topics and institutions for this page.

Published work

13 published item(s)

preprint2026arXiv

Lyapunov-Guided Self-Alignment: Test-Time Adaptation for Offline Safe Reinforcement Learning

Offline reinforcement learning (RL) agents often fail when deployed, as the gap between training datasets and real environments leads to unsafe behavior. To address this, we present SAS (Self-Alignment for Safety), a transformer-based framework that enables test-time adaptation in offline safe RL without retraining. In SAS, the main mechanism is self-alignment: at test time, the pretrained agent generates several imagined trajectories and selects those satisfying the Lyapunov condition. These feasible segments are then recycled as in-context prompts, allowing the agent to realign its behavior toward safety while avoiding parameter updates. In effect, SAS turns Lyapunov-guided imagination into control-invariant prompts, and its transformer architecture admits a hierarchical RL interpretation where prompting functions as Bayesian inference over latent skills. Across Safety Gymnasium and MuJoCo benchmarks, SAS consistently reduces cost and failure while maintaining or improving return.

0 citations0 reviewsNo code linkedOpen access
preprint2022arXiv

Gravitational Focusing of Wave Dark Matter

A massive astrophysical object deforms a local distribution of dark matter, resulting in a local overdensity of dark matter. This phenomenon is often referred to as gravitational focusing. In the solar system, the gravitational focusing due to the Sun induces modulations of dark matter signals on terrestrial experiments. We consider the gravitational focusing of a light bosonic dark matter with a mass of less than about 10 eV. The wave nature of such dark matter candidates leads to unique signatures both in the local overdensity and in the spectrum, both of which can be experimentally relevant. We provide a formalism that captures both the gravitational focusing and the stochasticity of wave dark matter, paying particular attention to the similarity and difference to particle dark matter. Distinctive patterns in the density contrast and spectrum are observed when the de Broglie wavelength of dark matter becomes comparable or less than the size of the system and/or when the velocity dispersion of dark matter is sufficiently small. While gravitational focusing effects generally remain at a few percent level for a relaxed halo dark matter component, they could be much larger for dark matter substructures. With a few well-motivated dark matter substructures, we investigate how each substructure responds to the gravitational potential of the Sun. The limit at which wave dark matter behaves similar to particle dark matter is also discussed.

0 citations0 reviewsNo code linkedOpen access
preprint2022arXiv

Platform-agnostic waveguide integration of high-speed photodetectors with evaporated tellurium thin films

Many attractive photonics platforms still lack integrated photodetectors due to inherent material incompatibilities and lack of process scalability, preventing their widespread deployment. Here we address the problem of scalably integrating photodetectors in a photonic platform-independent manner. Using a thermal evaporation and deposition technique developed for nanoelectronics, we show that tellurium (Te), a quasi-2D semi-conductive element, can be evaporated at low temperature directly onto photonic chips to form air-stable, high-responsivity, high-speed, ultrawide-band photodetectors. We demonstrate detection at visible, telecom, and mid-infrared wavelengths, a bandwidth of more than 40 GHz, and platform-independent scalable integration with photonic structures in silicon, silicon nitride and lithium niobate.

0 citations0 reviewsNo code linkedOpen access
preprint2021arXiv

Searching for Earth/Solar Axion Halos

We discuss the sensitivity of the present and near-future axion dark matter experiments to a halo of axions or axion-like particles gravitationally bound to the Earth or the Sun. The existence of such halos, assuming they are formed, renders a significant gain in the sensitivity of axion searches while satisfying all the present experimental bounds. The structure and coherence properties of these halos also imply novel signals, which can depend on the latitude or orientation of the detector. We demonstrate this by analysing the sensitivity of several distinct types of axion dark matter experiments.

0 citations0 reviewsNo code linkedOpen access
preprint2020arXiv

Probing the Relaxed Relaxion at the Luminosity and Precision Frontiers

Cosmological relaxation of the electroweak scale is an attractive scenario addressing the gauge hierarchy problem. Its main actor, the relaxion, is a light spin-zero field which dynamically relaxes the Higgs mass with respect to its natural large value. We show that the relaxion is generically stabilized at a special position in the field space, which leads to suppression of its mass and potentially unnatural values for the model's effective low-energy couplings. In particular, we find that the relaxion mixing with the Higgs can be several orders of magnitude above its naive naturalness bound. Low energy observers may thus find the relaxion theory being fine-tuned although the relaxion scenario itself is constructed in a technically natural way. More generally, we identify the lower and upper bounds on the mixing angle. We examine the experimental implications of the above observations at the luminosity and precision frontiers. A particular attention is given to the impressive ability of future nuclear clocks to search for rapidly oscillating scalar ultra-light dark matter, where the future projected sensitivity is presented.

0 citations0 reviewsNo code linkedOpen access
preprint2020arXiv

REST: Performance Improvement of a Black Box Model via RL-based Spatial Transformation

In recent years, deep neural networks (DNN) have become a highly active area of research, and shown remarkable achievements on a variety of computer vision tasks. DNNs, however, are known to often make overconfident yet incorrect predictions on out-of-distribution samples, which can be a major obstacle to real-world deployments because the training dataset is always limited compared to diverse real-world samples. Thus, it is fundamental to provide guarantees of robustness to the distribution shift between training and test time when we construct DNN models in practice. Moreover, in many cases, the deep learning models are deployed as black boxes and the performance has been already optimized for a training dataset, thus changing the black box itself can lead to performance degradation. We here study the robustness to the geometric transformations in a specific condition where the black-box image classifier is given. We propose an additional learner, \emph{REinforcement Spatial Transform learner (REST)}, that transforms the warped input data into samples regarded as in-distribution by the black-box models. Our work aims to improve the robustness by adding a REST module in front of any black boxes and training only the REST module without retraining the original black box model in an end-to-end manner, i.e. we try to convert the real-world data into training distribution which the performance of the black-box model is best suited for. We use a confidence score that is obtained from the black-box model to determine whether the transformed input is drawn from in-distribution. We empirically show that our method has an advantage in generalization to geometric transformations and sample efficiency.

0 citations0 reviewsNo code linkedOpen access
preprint2020arXiv

Searching for a solar relaxion/scalar with XENON1T and LUX

We consider liquid xenon dark matter detectors for searching a light scalar particle produced in the solar core, specifically one that couples to electrons. Through its interaction with the electrons, the scalar particle can be produced in the Sun, mainly through Bremsstrahlung process, and subsequently it is absorbed by liquid xenon atoms, leaving prompt scintillation light and ionization events. Using the latest experimental results of XENON1T and Large Underground Xenon, we place bounds on the coupling between electrons and a light scalar as $g_{ϕee} < 7 \times 10^{-15}$ from S1-only analysis, and as $g_{ϕee} < 2 \times 10^{-15}$ from S2-only analysis. These can be interpreted as bounds on the mixing angle with the Higgs, $\sin θ< 2 \times 10^{-9} \, \left(7 \times 10^{-10}\right)$, for the case of a relaxion that couples to the electrons via this mixing. The bounds are a factor few weaker than the strongest indirect bound inferred from stellar evolution considerations.

0 citations0 reviewsNo code linkedOpen access
preprint2019arXiv

Coherent relaxion dark matter

We show that relaxion, that addresses the hierarchy problem, can account for the observed dark matter (DM) relic density. The setup is similar to the case of axion DM models topped with a dynamical misalignment mechanism. After the reheating, when the temperature is well above the electroweak scale, the backreaction potential disappears and the relaxion is displaced from its vacuum. When the "wiggles" reappear the relaxion coherently oscillates around its minimum as in the case of vanilla axion DM models. We identify the parameter space such that the relaxion is retrapped leading to the standard cosmology. When the relaxion is lighter than $10^{-7}\,$eV, Hubble friction during radiation-domination is sufficiently strong for retrapping, and even minimal models are found to be viable. It also leads to a new constraint on relaxion models, as a certain region of their parameter space could lead to overabundant relaxion DM. Alternatively, even a larger parameter space exists when additional friction is obtained by particle production from additional coupling to an additional dark photon field. The phenomenology of this class of models is quite unique, as it implies that we are surrounded by a time-dependent axion-like field that due to relaxion-Higgs mixing implies time-dependent Higgs vacuum-expectation-value that lead to time-variation of all coupling constants of nature.

0 citations0 reviewsNo code linkedOpen access
preprint2019arXiv

Relaxion Stars and their detection via Atomic Physics

The cosmological relaxion can address the hierarchy problem, while its coherent oscillations can constitute dark matter in the present universe. We consider the possibility that the relaxion forms gravitationally bound objects that we denote as relaxion stars. The density of these stars would be higher than that of the local dark matter density, resulting in enhanced signals in table-top detectors, among others. Furthermore, we raise the possibility that these objects may be trapped by an external gravitational potential, such as that of the Earth or the Sun. This leads to formation of relaxion halos of even greater density. We discuss several interesting implications of relaxion halos, as well as detection strategies to probe them.

0 citations0 reviewsNo code linkedOpen access
preprint2016arXiv

750 GeV diphoton resonance and electric dipole moments

We examine the implication of the recently observed 750 GeV diphoton excess for the electric dipole moments of the neutron and electron. If the excess is due to a spin zero resonance which couples to photons and gluons through the loops of massive vector-like fermions, the resulting neutron electric dipole moment can be comparable to the present experimental bound if the CP-violating angle α in the underlying new physics is of O(10^{-1}). An electron EDM comparable to the present bound can be achieved through a mixing between the 750 GeV resonance and the Standard Model Higgs boson, if the mixing angle itself for an approximately pseudoscalar resonance, or the mixing angle times the CP-violating angle α for an approximately scalar resonance, is of O(10^{-3}). For the case that the 750 GeV resonance corresponds to a composite pseudo-Nambu-Goldstone boson formed by a QCD-like hypercolor dynamics confining at Λ_HC, the resulting neutron EDM can be estimated with α~ (750 GeV / Λ_HC)^2θ_HC, where θ_HC is the hypercolor vacuum angle.

0 citations0 reviewsNo code linkedOpen access
preprint2015arXiv

Primordial perturbations from dilaton-induced gauge fields

We study the primordial scalar and tensor perturbations in inflation scenario involving a spectator dilaton field. In our setup, the rolling spectator dilaton causes a tachyonic instability of gauge fields, leading to a copious production of gauge fields in the superhorizon regime, which generates additional scalar and tensor perturbations through gravitational interactions. Our prime concern is the possibility to enhance the tensor-to-scalar ratio $r$ relative to the standard result, while satisfying the observational constraints. To this end, we allow the dilaton field to be stabilized before the end of inflation, but after the CMB scales exit the horizon. We show that for the inflaton slow roll parameter $ε\gtrsim 10^{-3}$, the tensor-to-scalar ratio in our setup can be enhanced only by a factor of ${\cal O}(1)$ compared to the standard result. On the other hand, for smaller $ε$ corresponding to a lower inflation energy scale, a much bigger enhancement can be achieved, so that our setup can give rise to an observably large $r\gtrsim 10^{-2}$ even when $ε\ll 10^{-3}$. The tensor perturbation sourced by the spectator dilaton can have a strong scale dependence, and is generically red-tilted. We also discuss a specific model to realize our scenario, and identify the parameter region giving an observably large $r$ for relatively low inflation energy scales.

0 citations0 reviewsNo code linkedOpen access
preprint2014arXiv

Natural inflation with multiple sub-Planckian axions

We extend the Kim-Nilles-Peloso (KNP) alignment mechanism for natural inflation to models with $N>2$ axions, which obtains a super-Planckian effective axion decay constant $f_{\textrm{eff}}\gg M_{Pl}$ through an alignment of the anomaly coefficients of multiple axions having sub-Planckian fundamental decay constants $f_0\ll M_{Pl}$. The original version of the KNP mechanism realized with two axions requires that some of the anomaly coefficients should be of the order of $f_{\textrm{eff}}/f_0$, which would be uncomfortably large if $f_{\rm eff}/f_0 \gtrsim {\cal O}(100)$ as suggested by the recent BICEP2 results. We note that the KNP mechanism can be realized with the anomaly coefficients of $\mathcal{O}(1)$ if the number of axions $N$ is large as $N\ln N\gtrsim 2\ln (f_{\textrm{eff}}/f_0)$, in which case the effective decay constant can be enhanced as $f_{\rm eff}/f_0 \sim \sqrt{N !}\,n^{N-1}$ for $n$ denoting the typical size of the integer-valued anomaly coefficients. Comparing to the other multiple axion scenario, the N-flation scenario which requires $N \sim f_{\textrm{eff}}^2/f_0^2$, the KNP mechanism has a virtue of not invoking to a too large number of axions, although it requires a specific alignment of the anomaly coefficients, which can be achieved with a probability of ${\cal O}(f_0/f_{\rm eff})$ under a random choice of the anomaly coefficients. We also present a simple model realizing a multiple axion monodromy along the inflaton direction.

0 citations0 reviewsNo code linkedOpen access