Joshua S. Speagle

Generative artificial intelligence (AI) excels at producing complex data structures (text, images, videos) by learning patterns from training examples. Across scientific disciplines, researchers are now applying generative models to ‘inverse problems’ to directly predict hidden parameters from observed data along with measures of uncertainty. While these predictive or posterior-based methods can handle intractable likelihoods and large-scale studies, they can also produce biased or overconfident conclusions even without model misspecifications. We present a solution with Frequentist–Bayes (FreB), a mathematically rigorous protocol that reshapes AI-generated posterior probability distributions into (locally valid) confidence regions that consistently include true parameters with the expected probability, while achieving minimum size when training and target data align. We demonstrate FreB’s effectiveness by tackling diverse case studies in the physical sciences: identifying unknown sources under dataset shift, reconciling competing theoretical models, and mitigating selection bias and systematics in observational studies. By providing validity guarantees with interpretable diagnostics, FreB enables trustworthy scientific inference across fields where direct likelihood evaluation remains impossible or prohibitively expensive.

Observations of the early Universe ( z ≳ 4) with JWST reveal galaxy populations with a wide range of intrinsic luminosities and colors. Bursty star formation histories (SFHs), characterized by short-term fluctuations in the star formation rate (SFR), may explain this diversity, but constraining burst timescales and amplitudes in individual galaxies is challenging given degeneracies and sensitivity limits. We introduce a population-level simulation-based inference framework that recovers the power spectrum of SFR fluctuations by forward modeling galaxy populations and distributions of rest-frame UV to rest-frame optical spectral features sensitive to such timescales. We adopt a stochastic SFH model based on a power spectral density (PSD) formalism spanning 1 Myr–10 Gyr. Using simulated samples of N = 500 galaxies at z ∼ 4 with typical JWST/NIRSpec uncertainties, we demonstrate that (i) the PSD can be constrained sufficiently to distinguish between simulations (e.g., FIRE-2-like versus Illustris-like populations at >99% confidence for timescales < 100 Myr); (ii) simultaneously modeling stochastic fluctuations and the recent ( t _L < 500 Myr) SFH slope is essential, as secular trends mimic burstiness in common diagnostics; (iii) frequent, intense bursts impose an outshining limit, biasing the inference to underestimate burstiness; and (iv) the PSD can be inferred to 95% confidence in both smooth and bursty populations under consistent modeling assumptions. This framework establishes that large, uniformly selected spectroscopic samples can place quantitative constraints on SFR timescales and associated feedback processes. The biggest unexplored systematic effects are fixed stellar population synthesis models and the assumption of perfect dust corrections; we will address these in future work.

Zeeman observations provide the only direct probe of line-of-sight (LOS) magnetic fields in the interstellar medium. To evaluate their accuracy and limitations, we generate synthetic HI Zeeman spectra from magnetohydrodynamic simulations and idealized cloud models, and analyze the resulting Stokes I and V profiles using two complementary methods. Approach I uses the classical relation between Stokes V and dI/dν to estimate LOS-averaged magnetic fields, achieving an upper-limit relative error of 16% (half-width of 68.27% confidence interval) for a representative noise level of 0.014 K. Approach II applies Gaussian decomposition to Stokes I and V to estimate component-level magnetic fields, yielding a 13% relative error quantifying the same confidence range, reflecting the intrinsic uncertainty of such Zeeman estimates. Both approaches recover the original fields under uniform-field conditions and remain robust in turbulent environments. Approach I provides a simple and reliable LOS-averaged field estimate, while Approach II, although more complex, offers statistical insight into magnetic field variations along the LOS. We further show that joint fitting of Stokes I and V generally outperforms sequential fitting, particularly in the presence of attenuation. Increasing noise eight-fold produces a more modest rise in uncertainty, doubling to a 26% relative error, while substantial optical depth introduces only a minor additional contribution to the overall uncertainty. Applying these methods to FAST observations of the L1544 star-forming region, we confirm the previously reported LOS magnetic field strength, demonstrating the validity of Zeeman analysis in this benchmark core.

Observations of the early Universe (z > 4) with the James Webb Space Telescope reveal galaxy populations with a wide range of intrinsic luminosities and colors. Bursty star formation histories (SFHs), characterized by short-term fluctuations in the star formation rate (SFR), may explain this diversity, but constraining burst timescales and amplitudes in individual galaxies is challenging due to degeneracies and sensitivity limits. We introduce a population-level simulation-based inference framework that recovers the power and timescales of SFR fluctuations by forward-modeling galaxy populations and distributions of rest-UV to rest-optical spectral features sensitive to star formation timescales. We adopt a stochastic SFH model based on a power spectral density formalism spanning 1 Myr-10 Gyr. Using simulated samples of N=500 galaxies at z~4 with typical JWST/NIRSpec uncertainties, we demonstrate that: (i) the power of SFR fluctuations can be measured with sufficient precision to distinguish between simulations (e.g., FIRE-2-like vs. Illustris-like populations at >99% confidence for timescales < 100 Myr); (ii) simultaneously modeling stochastic fluctuations and the recent (t_L < 500 Myr) average SFH slope is essential, as secular trends otherwise mimic burstiness in common diagnostics; (iii) frequent, intense bursts impose an outshining limit, and bias inference toward underestimating burstiness due to the obscuration of long-timescale power; and (iv) the power of SFR fluctuations can be inferred to 95% confidence across all timescales in both smooth and bursty populations. This framework establishes a novel and robust method for placing quantitative constraints on the feedback physics regulating star formation using large, uniformly selected spectroscopic samples.

Recent Gaia-based young stellar association surveys have revealed dozens of low-mass populations that have, until recently, been too small or sparse to detect. These populations represent a largely unstudied demographic with unknown origins, and their relative isolation may minimize gravitational disruptions that impact traceback, making them compelling targets for dynamical studies. In this paper, we survey 15 of these isolated young associations for the first time: Andromeda South (SCYA-97), Aquila East, Aries South (SCYA-104), Cassiopeia East (SCYA-43), Canis Major North, Leo Central (SCYA-2), Leo East (SCYA-3), Theia 72, Ophiuchus Southeast, Scutum North (SCYA-70), Taurus-Orion 1 (TOR1), Theia 78, Vulpecula East (UPK 88), SCYA-54, and SCYA-79. By combining Gaia astrometry and photometry with new ground-based spectroscopic measurements, we assess the membership of each population, search for substructure, analyze their demographics, and compute ages. We find that the smallest populations in our sample contain M _⊙ of stellar mass, making them the smallest associations ever detected. Four host substantial substructure, including TOR1, where we discover TOR1B, a new 16 M _⊙ association with radial velocities inconsistent with an origin in the parent complex. Using PARSEC isochrones, we produce self-consistent ages for all populations supported by dynamical and lithium depletion ages, which range from 6.9 ± 0.5 Myr in TOR1A to 42.8 ± 2.4 in AndS. Our results provide the first detailed overview of the properties of these populations, characterizing a largely unknown category of young associations that may have an important role in tracing the processes that guide local star formation.

Spectral data reduction pipelines deal with a wide variety of challenges including masking cosmic rays, calibrating wavelength solutions, and estimating background noise while trying to remain model-agnostic. Traditional methods rely on hardware-specific code or pre-calculated stellar model templates to solve this problem, making them model-dependent and not suitable for large datasets that may contain new classes of objects. To solve this problem, we present a flexible, data-driven method: the GausSian PIxelwise Conditional Estimator (GSPICE) that models an ensemble of spectra as a multivariate Gaussian and estimates the expected value and expected variance of each pixel in each spectrum conditional on others. GSPICE compares observed fluxes and errors to its own flux and error estimates to reveal outliers, which then can be completely masked or replaced by their estimates. We apply GSPICE to 3.9 million stellar spectra from the LAMOST survey, and show that variations of the method can directly identify and correct both individual pixel-level outliers (e.g., from cosmic ray hits) as well as extended systematic features (e.g., from incorrect wavelength calibrations), while still providing a novel characterization of the true per-pixel measurement uncertainties. We also demonstrate how GSPICE can take advantage of data partitioning with an application to diffuse interstellar bands. Implementations of GSPICE in both Python and IDL can be found here http://github.com/dfink/gspice.

This decade has seen the first measurements of extrasolar planetary obliquities, characterizing how an exoplanet's spin axis is oriented relative to its orbital axis. These measurements are enabled by combining projected rotational velocities, planetary rotation periods, and astrometric orbits for directly-imaged super-Jupiters. This approach constrains both the spin axis and orbital inclination relative to the line of sight, allowing obliquity measurements for individual systems and offering new insights into their formation. To test whether these super-Jupiters form more like scaled-up planets or scaled-down stars, we develop a hierarchical Bayesian framework to infer their population-level obliquity distribution. Using a single-parameter Fisher distribution, we compare two models: a planet-like formation scenario ($κ=5$) predicting moderate alignment, versus a brown dwarf-like formation scenario ($κ=0$) predicting isotropic obliquities. Based on a sample of four young super-Jupiter systems, we find early evidence favoring the isotropic case with a Bayes factor of 15, consistent with turbulent fragmentation.

We present a deep, high-angular-resolution 3D dust map of the southern Galactic plane over 239° < l < 6° and ∣ b∣ < 10° built on photometry from the DECaPS2 survey, in combination with photometry from VISTA Variables in the Via Lactea, the Two Micron All Sky Survey, and “Unofficial” Wide-field Infrared Survey Explorer and parallaxes from Gaia Data Release 3 where available. To construct the map, we first infer the distance, extinction, and stellar types of over 700 million stars using the brutus stellar inference framework with a set of theoretical MESA Isochrone and Stellar Tracks (MIST) stellar models. Our resultant 3D dust map has an angular resolution of 1 ′ , roughly an order of magnitude finer than existing 3D dust maps and comparable to the angular resolution of the Herschel 2D dust emission maps. We detect complexes at the range of distances associated with the Sagittarius-Carina and Scutum-Centaurus arms in the fourth quadrant, as well as more distant structures out to a maximum reliable distance of d ≈ 10 kpc from the Sun. The map is sensitive up to a maximum extinction of roughly A _V ≈ 12 mag. We publicly release both the stellar catalog and the 3D dust map, the latter of which can easily be queried via the Python package dustmaps. When combined with the existing Bayestar19 3D dust map of the northern sky, the DECaPS 3D dust map fills in the missing piece of the Galactic plane, enabling extinction corrections over the entire disk ∣b∣ < 10°. Our map serves as a pathfinder for the future of 3D dust mapping in the era of LSST and Roman, targeting regimes accessible with deep optical and near-infrared photometry but often inaccessible with Gaia.