Daily Summary for 2026-11-09
daily
Transcript
Introduction to the show: ident: Astrophysics. The day in astrophysics: observations, models and what the data forces.
Vera: It's the ninth of November, twenty twenty-six, and this is the day's research.
Jocelyn: 17 new papers came out today.
Vera: I'm Vera.
Jocelyn: And I'm Jocelyn, here with Subrahmanyan.
Vera: We'll take the day in one pass, then pull out the papers we're staying with.
Jocelyn: Start us off with Latin hypercube sampling to reduce the variance in Monte Carlo integration.
The day's research: Vera: Today we focus on cleaning noisy images for better measurements.
Jocelyn: FiDeR uses finite-iteration deconvolution and reconvolution denoising to reduce noise while keeping the original image structure intact.
Subrahmanyan: That technique showed superior photometric accuracy, reducing relative mean deviation to below one percent for isolated sources in simulations.
Vera: FiDeR shows gains over traditional denoisers when applied to real data from instruments like the James Webb Space Telescope Near-Infrared Camera.
Jocelyn: This approach builds on prior work using Richardson--Lucy iterations, but FiDeR introduces a novel filtered image deconvolution and noise backfilling algorithm.
Subrahmanyan: This algorithm integrates a low-pass prefiltering stage using a Kalman filter before reintegrating filtered noise components after the RL steps.
Vera: This combination effectively suppresses artifacts while preserving flux integrity, leading to lower core RMSE in JWST observations compared to methods like BM3D.
Jocelyn: Another related study introduced FiDeR, which calibrates the point-source response of a finite RL reconstruction and derives a reconvolution kernel matched to the observational PSF.
Subrahmanyan: This method achieves high signal-to-noise ratios and lower aperture-flux MAEs than BM3D in simulated fields.
Vera: We are also looking at how statistical sampling can improve the reliability of these integrations.
Jocelyn: A paper on Latin hypercube sampling shows that using orthogonal arrays can reduce the variance in Monte Carlo integration by orders of magnitude when the integrand is dominated by additive contributions or pairwise interactions between coordinates.
Subrahmanyan: This is relevant for gravitational-wave astronomy likelihood estimates.
Vera: The most significant finding relates to how polarization cross-talk from non-stationary sources of image jitter can be modeled.
Jocelyn: This modeling is crucial because accurately characterizing this noise affects our ability to interpret polarization measurements in astrophysical data.
Subrahmanyan: The work shows that a generalized formalism converges with previous models when the image jitter becomes stationary, but it highlights distinct differences when dealing with non-stationary processes like those induced by atmospheric seeing.
Vera: Understanding these jitter effects directly impacts the fidelity of polarization maps, which are essential for probing magnetic field geometry in high-energy sources.
Jocelyn: The study compares predictions based on a power spectral density derived from a circulant auto-correlation function against the extended modeling framework developed in Polarization cross-talk from non-stationary sources of image jitter.
Subrahmanyan: The study on deconvolving IXPE polarization maps of the Crab Nebula is also vital because it demonstrates how applying a maximum a posteriori framework can yield sharper spatial structures in reconstructed images.
Vera: This result shows a statistically significant relationship between magnetic field orientation and intensity gradients in jet regions, suggesting that polarization-resolved deconvolution enhances scientific interpretability by revealing structures hidden in basic analyses.
Jocelyn: Furthermore, the detection of differential chromatic refraction with LSST camera data provides new constraints on photometric redshifts by encoding information about the source's spectral energy distribution within the characteristic wavelength lambdaDCR.
Subrahmanyan: This is a key piece of information because it helps break degeneracies when selecting and characterizing extragalactic samples from large surveys.
Vera: This DCR effect is particularly interesting because for quasars, strong emission lines skew the mean astrometric deflection angle in the u band, while for red galaxies, the r band lambdaDCR shows a clear shift around redshift zero point four.
Jocelyn: This suggests that DCR offers complementary information to broadband colors when estimating photometric redshifts.
Subrahmanyan: Detection and excision of radio-frequency interference in radio arrays develops models to suppress radio frequency interference in antenna observations using various array characteristics.
Vera: From Rock Tumbling to Telescope Mirror Grinding explains how optical surfaces can become perfectly smooth through random rubbing processes by analyzing material removal dynamics.
Jocelyn: GALI Localization of Off-Grid GRB Directions compares different calibration meshes for accurately localizing gamma-ray bursts using count rate modulations in a scintillator array.
Subrahmanyan: Tensor Networks for Block Adaptive Thermal Radiation Transport introduces a method combining adaptive mesh refinement with tensor networks to efficiently solve complex thermal radiation transport problems.
Vera: Latin hypercube sampling to reduce the variance in Monte Carlo integration provides a practical guide and application of Latin hypercube sampling to reduce variance in Monte Carlo integration for gravitational-wave astronomy likelihoods.
Jocelyn: Aperture Photometric Accuracy of Point-spread-function-deconvolved Astronomical Images introduces a new method using Kalman filtering and Richardson--Lucy iterations to improve photometric accuracy in astronomical images.
Subrahmanyan: The infrared unmodulated pyramid wavefront sensor investigates an infrared version of a pyramid wavefront sensor that offers faster operation and improved performance for extreme adaptive optics.
Vera: FiDeR Denoising Astronomical Images at the Observational Resolution presents a method called FiDeR to reduce noise in astronomical images while preserving the input point-spread function.
Jocelyn: Polarization cross-talk from non-stationary sources of image jitter generalizes a formulation to estimate polarization cross-talk caused by non-stationary, random image jitter.
Subrahmanyan: Preliminary Goniopolarimetric Simulations for Space Weather Radio SWARA Polarimeter explores preliminary simulations for a tri-monopole antenna configuration to develop an early warning polarimeter for space weather.
Vera: Inertial sub-range detection and characterization for ground-layer turbulence at new extremely large telescope sites in Chile uses detrended fluctuation analysis to identify the inertial sub-range of atmospheric turbulence at large telescopes.
Jocelyn: Deconvolved IXPE Polarization Maps of the Crab Nebula in a Maximum-a-Posteriori Framework uses maximum a posteriori estimation to reconstruct sharper polarization maps and reveal magnetic field structure in the Crab Nebula.
Subrahmanyan: Mitigating scanning-induced systematics in Hipparcos and implications for Gaia astrometry proposes a method to reduce scan geometry dependence when searching for periodic signals in space-based astrometry data.
Vera: Why a Multi-Cycle Treasury Program is Essential to Maximize JWST's Scientific Contributions to Galaxy Formation argues that coordinated, large surveys are necessary for JWST to achieve its full scientific potential in understanding galaxy evolution.
Jocelyn: On the importance of medium-resolution continuum spectroscopy for JWST's extragalactic legacy advocates for deep medium-resolution spectroscopy on JWST to constrain key physical properties of high-redshift galaxies.
Subrahmanyan: Detection of Differential Chromatic Refraction with LSST Camera data shows how differential chromatic refraction in LSST camera data can provide information about the wavelength distribution of light in astronomical sources.
Vera: Towards Genuine Coexistence models the equivalent power flux density of satellite constellations to determine protection limits for radio astronomy and geodetic VLBI at 1-14 GHz from satellite constellations.
Jocelyn: Per-Satellite Emission and Radiation Limits to Protect Radio Astronomy and Geodetic VLBI at 1-14 GHz from Satellite Constellations models the equivalent power flux density of satellite constellations to determine protection limits for radio astronomy and geodetic VLBI at 1-14 GHz from satellite constellations.
Lucky paper: 2611.04045: Vera: So we're looking at Latin hypercube sampling to reduce the variance in Monte Carlo integration, which is a practical guide and application to gravitational-wave astronomy.
Jocelyn: It sounds like this paper is tackling the problem of high-dimensional integrals in statistical inference where you need reliable results.
Subrahmanyan: The paper explains that when likelihoods involve high-dimensional integrals, we often use Monte Carlo summation, and controlling that numerical variance is really crucial for getting reliable inferences.
Vera: It introduces Latin hypercube sampling as a space-filling strategy for drawing randomized multidimensional samples with uniform stratification in each one-dimensional marginal.
Jocelyn: And they mention orthogonal array Latin hypercube sampling, or OA-LHS, extends this stratification to higher dimensional marginals.
Subrahmanyan: They show that under appropriate regularity conditions, the asymptotic variance of an LHS based Monte Carlo estimator is mathematically guaranteed to be no greater than that of a standard Monte Carlo estimator.
Vera: The mechanism for Latin hypercube sampling involves drawing N points in the unit hypercube in d dimensions, where they divide each dimension into N disjoint bins and select one bin uniformly at random in each dimension independently across dimensions.
Jocelyn: That construction formula they give is ui j equals pi j i plus Vij over N, where Vij is drawn from a uniform distribution between zero and one.
Subrahmanyan: They then decompose the variance of an LHS based Monte Carlo estimator into additive and interaction terms, showing the asymptotic variance for LHS is equal to one over N times sigma squared minus X j sigma squared j.
Vera: The application here is gravitational-wave astronomy where hierarchical Bayesian statistics is used to estimate population properties of merging black holes and neutron stars.
Jocelyn: They are investigating the effect of LHS and OA-LHS on the variance of a selection function estimate, which they define as xi lambda equals Z ppop theta lambda pinj theta i pinj p det theta i.
Subrahmanyan: They found that for this application, the ratio R equal to Var cLHS over Var cMC follows an F distribution with R minus one and R minus one degrees of freedom.
Vera: The practical results show that for the selection function estimates shown in Figure four the ratios RLHS is zero point nine six and ROA-LHS is zero point eight nine when N equals nine hundred forty thousand injected binaries.
Jocelyn: They conclude that they found no statistically significant difference between the variances of these three sampling methods for this specific application, which is interesting given the theoretical bounds they laid out earlier.
Subrahmanyan: The authors recommend drawing the next set of LIGO Virgo injections using OA-LHS because the computational cost to construct these designs is negligible relative to evaluating gravitational-wave injections.
Vera: The paper summarizes that strength two orthogonal arrays can reduce variance by orders of magnitude when the integrand is dominated by additive contributions or pairwise interactions between coordinates.
Jocelyn: So, even though they found no statistically significant difference in this specific test, the potential for massive variance reduction with OA-LHS when those interaction terms are dominant is what makes this method useful.
Subrahmanyan: It's a practical guide because it shows how these space-filling designs can reduce the asymptotic variance of the resulting estimator.
Lucky paper: 2611.01657: Vera: So, we're looking at FiDeR, which is this method for denoising astronomical images while trying to keep the original point-spread function intact.
Jocelyn: That method uses finite iteration of Richardson--Lucy iterations on a nonnegative source-plus-background image, and it derives a reconvolution kernel matched to the observational point-spread function.
Subrahmanyan: It introduces this calibration step where you measure the response of a calibration point source processed with the same settings to derive that reconvolution kernel.
Vera: They compared FiDeR against six other denoisers using simulated fields, and the results showed it reaching peak signal-to-noise ratios of forty-six point seven zero one and forty-four point eight five zero dB at backgrounds of ten and twenty counts per pixel.
Jocelyn: And when looking at aperture flux mean absolute errors, FiDeR achieved values of zero point nine six percent and one point zero six percent, which was better than BMthree dee's values of one point zero four percent and one point one eight percent.
Subrahmanyan: The paper also looked at JWST observations, where FiDeR reduced the core root mean square error to sixteen point five nine eight counts, down from thirty-five point three six six for Raw data and BMthree dee's thirty-five point three zero four counts.
Vera: It seems like it really helps when you look at those specific JWST observations, especially in the stellar aperture where it gets a very low error of zero point one two three zero plus or minus zero point zero one five seven MJy per square degree.
Jocelyn: That is quite a difference compared to BMthree dee's performance in that same region, and they also showed that FiDeR has the highest combined structural similarity index of zero point nine nine nine eight four eight five when compared to BMthree dee's zero point nine nine nine seven eight zero six.
Subrahmanyan: The method is based on a Poisson negative log likelihood update where the RL update is x sub j plus one equals x sub j times H T h y H x sub j i, starting with zero initial iteration budget of two hundred or one thousand.
Vera: That finite iteration budget of two hundred or one thousand sounds like a concrete thing you have to decide when you run the reconstruction.
Jocelyn: They also looked at how the method handles mixed fields, where FiDeR still managed to keep those high signal-to-noise ratios at forty-six point seven zero one and forty-four point eight five zero dB even with backgrounds of ten or twenty counts per pixel.
Subrahmanyan: The paper also provided some diagnostic definitions, like using an eleven by eleven Gaussian window with a standard deviation of one point five pixels to measure the residual difference between the output image and the reference image.
Vera: That structural similarity index measurement sounds like it's how they quantify how much spatial structure FiDeR is actually preserving compared to what's in the original data.
Jocelyn: They also discussed that FiDeR achieves substantial noise reduction with little change in the measured point-spread function across several scenes, which is important for its application in repeated exposures before UPDC deconvolution.
Subrahmanyan: If you look at the implementation details, they used BMthree dee version four point zero.three and its dependency on BM4D version four point two.five for running the process.
Vera: It seems like a practical tool here because it's being used in actual observations from NIRCam and MIRI instruments to see those error reductions in action, not just simulations.
Jocelyn: And they also looked at the MIRI observations, where FiDeR reduced the full image root mean square error from zero point one four six one to zero point one one one six MJy per square degree and boosted the signal-to-noise ratio from fifty-nine point nine nine one to sixty-two point two one five dB.
Subrahmanyan: For the stellar aperture specifically, that error dropped down to zero point one two three zero plus or minus zero point zero one five seven MJy per square degree, which is a very low value for those conditions.
Vera: It really shows how much better it performs when you're focusing on the stellar regions in those infrared data sets compared to older methods like BMthree dee.
Jocelyn: The conclusion emphasizes that these properties can benefit machine learning inputs and morphological measurements, provided you do end-to-end tests that account for the covariance of the processed pixels.
Subrahmanyan: And they also touched on how this relates to the paper on Latin hypercube sampling, showing it can reduce variance in Monte Carlo integration by orders of magnitude when dealing with additive contributions.
Vera: It seems like they're connecting these different statistical improvements—the deconvolution and the sampling—to make the entire process more reliable for complex calculations.
Lucky paper: 2611.08300: Vera: Okay, so we're looking at this paper now, it's titled Polarization cross-talk from non-stationary sources of image jitter.
Jocelyn: It generalizes a previous formulation for estimating polarization cross-talk induced by atmospheric seeing to the case of image jitter that has to be described as a non-stationary and generally non-isotropic random process.
Subrahmanyan: The mean square error on the demodulated Stokes parameters is still given by Equation six but the matrix On is now determined by the two frequency power spectral density P alpha beta omega omega prime of the two time autocorrelation function of the jitter, through Equation seventeen or Equation twenty for a stepped modulator.
Vera: So, what's this generalization actually doing in practice?
Jocelyn: The paper extends a formalism from Casini et al. two thousand twelve to image motion that is not stationary and possibly non-isotropic.
Subrahmanyan: The key change involves interpreting the Stokes vector and its gradient in Equation eight of Paper I as the unperturbed Stokes vector and its gradient at the fixed position x zero, rather than ensemble averages.
Vera: So, what does that mean for how we think about image motion?
Jocelyn: This means the image motion is no longer required to have zero mean; instead, the two time autocorrelation function of Equation twelve corresponds to the second order moment of that displacement from x zero.
Subrahmanyan: The resulting error estimate yields the mean square error of the inferred Stokes parameters, which includes any systematic drift of the image.
Vera: That sounds like it accounts for things we usually ignore, like systematic drift across a long exposure.
Jocelyn: Exactly, and then they talk about modeling non-stationarity in their study.
Subrahmanyan: Non-stationarity in this study arises from contributions to image jitter that are not invariant under a time shift of the polarization modulation sequence, such as pointing drifts, mechanical and thermal transients, and periodic components.
Vera: So they're talking about things like telescope drifting or thermal changes messing up the polarization signal.
Jocelyn: The expectation values introduced in Section two have to be interpreted as direct unperturbed Stokes vector and its gradient at the fixed position x zero.
Subrahmanyan: For a generally non-stationary random image motion, the expectation value in the integral of Equation eleven corresponds to the two time autocorrelation function Gamma alpha beta t t prime.
Vera: And then they characterize that using a two dimensional power spectral density P alpha beta omega omega prime obtained via a two dimensional Fourier transform.
Jocelyn: In the limit of a stationary random process, the two dimensional power spectral density is non vanishing only along the diagonal, P alpha beta omega omega prime equals two pi delta of omega minus omega prime S alpha beta omega, and Equation seventeen reduces identically to Equation twenty-one recovering the results of Paper I.
Subrahmanyan: Numerical tests using a circulant two dimensional autocorrelation show that cross-talk estimates from the full two dimensional model and from the one dimensional model applied to the diagonal power spectral density agree within typical numerical noise.
Vera: So they confirm that when things are stationary, the simpler one dimensional model works fine for estimating cross-talk.
Jocelyn: However, when power characteristics violate stationarity, significant off diagonal structure in the two dimensional power spectral density may include a contribution to cross-talk that is entirely missed by the one dimensional power spectral density model.
Subrahmanyan: The paper concludes that the one dimensional power spectral density does not carry enough information to recover the possible non stationary character of the associated random process.
Vera: So, what's the practical implication for someone actually dealing with real data?
Jocelyn: The application of this formalism fundamentally relies on the availability of the two time autocorrelation function, or which is equivalent, of the time series of the jitter from which Gamma alpha beta t t prime can be estimated.
Subrahmanyan: An alternative is to utilize realistic physical models to generate an ensemble of statistically independent realizations and hence Gamma alpha beta t t prime.
Vera: The simpler stationary model can still be useful to correctly identify the critical rates of the modulation cycle and provide an approximate estimation of their cross-talk effects.
Jocelyn: That means even if we don't have the full time series, we can use a stationary approximation to get a rough idea of those cross-talk effects.
Lucky paper: 2611.01820: Vera: So we're looking at the paper Deconvolved IXPE Polarization Maps of the Crab Nebula in a Maximum-a-Posteriori Framework, which uses maximum a posteriori estimation to reconstruct sharper spatial structures from the first IXPE observation of the Crab Nebula.
Jocelyn: That method formulates the reconstruction as an inverse problem where you discretize both the detector space for observed data and sky polarization space for unknown intensity.
Subrahmanyan: The analysis focuses on one specific observation, ObsID one million one thousand ninety-nine with an on-source exposure of ninety-two point three ks.
Vera: They model the sky emission as a sum of two components: the nebular component Is and the point-source component I(p) point,s.
Jocelyn: The expected counts are written as E(c,p)t equals X N s=one m(c) t,s Is plus I(p) point,s dp.
Subrahmanyan: They optimize the hyperparameter set mu using K-fold cross-validation to minimize the cross-validation score RMSE of mu.
Vera: They use a one standard error heuristic to reduce the risk of overfitting to training data when finding the optimal hyperparameter set mu star using Bayesian optimization.
Jocelyn: The results show polarization-resolved maps with sharper spatial structure than what you would get from a basic cube based analysis.
Subrahmanyan: In the jet body to tip subregion S, the reconstructed magnetic field orientation shows a statistically significant relation to the local intensity-gradient orientation.
Vera: They report that the weighted mean resultant length of angular differences R bar w is larger than what you would expect under null hypotheses with no polarization and randomized polarization angles at p equals zero point zero zero one two and zero point zero zero one eight, respectively.
Jocelyn: That means the recovered polarization pattern at jet regions is unlikely to be explained by statistical fluctuations, which is a strong point for them.
Subrahmanyan: The conclusion states that polarization-resolved deconvolution with smoothness priors enhances scientific interpretability by revealing spatial structures that are difficult to isolate in the basic analysis.
Vera: They also mention that the monochromatic approximation does not account for these differences and may cause spatially varying systematic errors in the reconstructed polarization degree.
Jocelyn: And they noted that this same effect increases the empirical MDP even in unpolarized Monte Carlo realizations, which is a bit concerning for consistency.
Subrahmanyan: They also point out that since their method is formulated in polarization-angle count space, the Stokes based polarization leakage treatment from Wong et al. two thousand twenty-four cannot be applied directly.
Vera: So how do they handle that leakage issue within their framework?
Jocelyn: Instead of using a Stokes based treatment, they incorporate the polarization leakage into the simulated detector response used to construct the response matrix instead.
Subrahmanyan: They also discuss varying the polarization angle binning according to local photon counts might improve reconstruction stability.
Vera: Thinking about what this means for us who only listen, it suggests that when you look at complex high energy data, just looking at the basic image structure isn't enough to see the true magnetic field geometry.
Jocelyn: It's about using that maximum a posteriori framework to push through the noise and reveal structures that are otherwise hidden by simple analysis.
Subrahmanyan: The paper was supported by a JSPS KAKENHI Grant Number JP19H05609, which is helpful context for the work.
Vera: Okay, so to wrap up this segment, we've heard about how they used MAP estimation to get sharper maps and found a significant perpendicular relationship between magnetic field orientation and intensity gradients in the jet region S.
Jocelyn: And while they show this relationship holds across region S, it's important to keep in mind that the monochromatic approximation might still introduce spatially varying systematic errors in the polarization degree.
Subrahmanyan: They also suggest that comparing the polarization pattern they recovered from their deconvolved jet map with theoretical models will be an interesting direction for future study.
Lucky paper: 2611.07027: Vera: Okay, we're moving into the paper titled Why a Multi-Cycle Treasury Program is Essential to Maximize JWST's Scientific Contributions to Galaxy Formation.
Jocelyn: This paper is really about how coordinating massive, multi-cycle programs can give JWST the statistical control it needs to answer those big questions about galaxy formation.
Subrahmanyan: The authors bring together a lot of different institutions, from UCLA and the University of Edinburgh all the way to INAF in Rome.
Vera: So what's the core problem they are addressing right now?
Jocelyn: They point out that current JWST programs have this issue where they don't have a uniform, statistically controlled legacy dataset yet.
Subrahmanyan: They argue that to get a quantitative physical understanding of early galaxy formation, we need that kind of uniform data across multiple cycles.
Vera: So they are suggesting a multi-cycle Treasury Program to fix the current limitations?
Jocelyn: Exactly, and they show that the largest programs like NEXUS, VENUS, and COSMOS-Web are already hitting a practical ceiling on what is approved for scheduling.
Subrahmanyan: They estimate that the science described in this paper demands time allocations of about one thousand to two thousand hours.
Vera: What's the real scientific imperative driving all this need for coordination?
Jocelyn: The two primary goals are discovering and characterizing the first stars and galaxies, and understanding how those objects evolved into the galaxies we see today.
Subrahmanyan: They link these questions—like when the first galaxies emerged from the cosmic dark ages—to everything else, including how black holes formed and how heavy elements for life were made.
Vera: And they are focusing on scaling relations as a way to encode these complex processes in the observable properties of galaxies.
Jocelyn: They mention that current empirical extrapolations suggest that the cosmic star formation rate density and UV-bright galaxy number density would continue to decline towards even higher redshifts.
Subrahmanyan: The paper highlights how a comprehensive program would optimize survey design to use resources like the MSA with uniform selection functions, which is crucial for statistical studies.
Vera: Meng, from an engineering standpoint, what does that uniformity actually translate to when you're running models? What breaks at scale?
Meng: If you don't have uniform selection functions, your statistical studies get messy because the sample isn't representative of the whole population.
Jocelyn: They argue that heterogeneous criteria across different programs limit the construction of statistically complete and uniform samples right now.
Subrahmanyan: A comprehensive program would naturally yield high-level data products and catalogs that you can use for comparison with theory, which is a big deal for validation.
Vera: It sounds like they are also looking ahead to how this coordinated effort will guide future missions, like ALMA or Euclid.
Meng: It's about maximizing synergy with those upcoming facilities, identifying high-value science targets for the next generation of instruments.
Jocelyn: The paper stresses that these programs are time-critical because most instruments are still at peak performance right now.
Subrahmanyan: They also point out that some modes, like longer wavelength MIRI or NIRSpec MSA growing stuck shutters, are already at risk of reduced efficiency.
Vera: So the call isn't just about getting more data, it’s about strategic resource utilization across multiple missions.
Jocelyn: Right. They show that this multi-cycle strategy provides the scale and continuity needed to address those interconnected science goals over time.
More episodes
- 2607.17929-gevolution 2.0: GPU-accelerated relativistic N-body simulations for cosmology
- The day in conversation for 2026-11-19
- Daily Summary for 2026-11-19
- 2509.02657-On the synergetic use of Ariel and JWST for exoplanet atmospheric science
- 2609.10042-Modelling Fourier filtering wavefront sensors for PSD-based methods: the AOPERA tool
- 2511.05639-Performance of the AstroPix Prototype Module for the Barrel Imaging Calorimeter at the ePIC Detector and in Space-Based Payloads
- 2610.19909-RPCA: a Resolved PSF reconstruction method for undersampled images
- 2512.03440-An Analysis of LIGO Glitches Using t-SNE During the First Part of the Fourth LIGO-Virgo-KAGRA Observing Run
- 2608.10284-Lunar Reflective Interferometry
- 2610.16633-Institutional broadening of the Japanese astronomy and astrophysics research community during 1996-2025