Searcharxiv⌕ Search

arXiv subjects

Jameson Dong

Publications and source records attributed to Jameson Dong.

6 recordsLinked to original sources

A High Precision Transformer model for precessing binary black-hole merger waveforms calibrated to numerical relativity

We present a Transformer model for merger-centered gravitational waveforms from precessing binary black holes. Given the mass ratio and two three-dimensional spin vectors, the network predicts quintic B-spline coefficients for all 21 signed spin-weighted spherical-harmonic strain modes through $\ell=4$ on $[-1100,80]M$ ($M$ being the total binary mass), enforcing binary-exchange and equatorial-reflection equivariance and yielding $C^4$ waveforms in time with automatic differentiation. The model is pretrained on numerical-relativity (NR) waveforms from the SXS and RIT catalogues together with synthetic NRSur7dq4, SEOBNRv5PHM, and perturbative waveforms, and fine-tuned on SXS strain data. On 109 SXS evaluation waveforms (106 physical systems), it achieves a mean balanced root-mean-square strain error of 0.0279 and a mean finite-window detector mismatch at $120\,M_\odot$ of 0.1190%.

gr-qc↗

ENFR: Evidence Networks with Flow-Matched References for Amortized Simulation-Based Model Inference

The evidence, $\mathcal{Z} = p(x | \mathcal{M})$, of a data-generating model $\mathcal{M}$ is a powerful tool in Bayesian frameworks: it enables model inference, model misspecification tests, and goodness-of-fit assessment as an alternative to standard discrepancy statistics, e.g. $χ^2$. While versatile, the evidence is notoriously difficult to compute, even when the likelihood is available. In both likelihood-based and simulation-based inference, estimating $\mathcal{Z}$ provides an immediate solution to high-dimensional goodness-of-fit or model misspecification tests: using $\mathcal{Z}(x)$ as the test statistic induces a minimum-volume level-$α$ test. Such tests are essential to assess model fidelity, which is necessary to trust simulation-based parameter inference. Existing learned approaches largely target Bayes factors, which only measure evidence ratios. Instead, we provide accurate, fast, and amortized estimates of the absolute evidence. We present ENFR: Evidence Networks with Flow-matched References. Building on the Evidence Network framework of Jeffrey & Wandelt (2024), the proposed method first learns Bayes factors with respect to a flow-matched reference, from which the absolute evidence is recovered. We show improvement in estimation of $\mathcal{Z}$ over the flow-based baseline in all applications considered and the best performance of all density estimation methods considered in our benchmarks. We furthermore introduce and provide the proof for an architecture that jointly learns the individual evidences of a set of models $\{\mathcal{M}_i\}$. We demonstrate the performance and scaling behavior of these approaches on data from a variety of scientific applications.

stat.ML↗

CoRe: Co-Evolving Reward Models for Mitigating Latent Reward Hacking in Video Diffusion Models

Latent reward models (LRMs) enable efficient alignment of video diffusion models by scoring intermediate states directly in latent space. However, we find that optimizing against a fixed latent reward rapidly leads to latent reward hacking: the predicted reward stays high while perceptual and motion quality deteriorate. Our analysis identifies distributional escape as the central cause: within a few hundred updates, the generator moves beyond the reward model's training support, where its scores no longer reflect video quality. Based on this insight, we introduce CoRe, a co-evolving reward framework that treats latent-space alignment as a dynamic interaction between the generator and the reward model. Rather than optimizing against a stationary proxy, CoRe continually refits the reward model on the generator's current samples while anchoring it to real-video preferences, so the generator cannot gain reward by drifting away from the data. On Wan2.1-T2V-1.3B, experiments show that CoRe consistently improves generation quality over both the pretrained model and prior alignment methods, while avoiding the quality collapse of fixed-reward optimization.

cs.AI↗

Next Generation Accretion Disk Reflection Model: High-Density Plasma Effects

Luminous accretion disks around black holes are expected to have densities of $\sim 10^{15-22}\,$cm$^{-3}$, which are high enough such that plasma physics effects become important. Many of these effects have been traditionally neglected in the calculation of atomic parameters, and therefore from photoionization models, and ultimately also from X-ray reflection models. In this paper, we describe updates to the atomic rates used by the XSTAR code, which is in turn part of the XILLVER disk reflection model. We discuss the effect of adding necessary high density corrections into the XILLVER code. Specifically, we find that the change of recombination rates play an important role, dominating the differences between model versions. With synthetic spectra, we show that even in a highly ionized state, high density slabs can produce strong iron ($\sim$6.5-9$\,$keV) and oxygen ($\sim0.6-0.8\,$keV) resonance features. The significant iron emission could address the problem of the supersolar iron abundances found in some sources.

astro-ph.HE↗

X-ray Reflection from the Plunging Region of Black Hole Accretion Disks

Accretion around black holes is very often characterized by distinctive X-ray reflection features (mostly, iron inner-shell transitions), which arise due to the primary radiation being reprocessed by a dense and relatively colder medium, such as an accretion disk. Most reflection modeling assume that emission stops at the inner-most stable circular orbit (ISCO), and that for smaller radii - in the plunging region - the density drops and the accretion flow is far too ionized for efficient line production. We investigate the spectral features of the reflection in the plunging regions of optically-thick and geometrically-thin accretion disks around black holes. We show that for cases in which the density profile is considered constant (as expected in highly magnetized flows), or in cases in which the disk density is high enough such that the ionization still allows line formation within the ISCO, there is a significant modification of the observed reflected spectrum. Consistent with previous studies, we found that the impact of the radiation reprocessed in the plunging region is stronger the lower the black hole spin, when the plunging region subtends a larger area. Likewise, as for the case of standard reflection modeling, the relativistic broadening of the iron line is more pronounced at low inclination, whereas the blueshift and relativistic beaming effect is dominant at high inclination. We also tested the effects of various prescriptions of the stress at the ISCO radius on the reflection spectrum, and found that several of these cases appear to show line profiles distinct enough to be distinguishable with reasonably good quality observational data.

astro-ph.HE↗

Blandford-Znajek Process in Quadratic Gravity

The Blandford-Znajek process, which uses a magnetized plasma to extract energy from a rotating black hole, is one of the leading candidates for powering relativistic jets. In this work, we investigate the Blandford-Znajek process in two well-motivated quadratic gravity theories: scalar Gauss-Bonnet and dynamical Chern-Simons gravity. We solve analytically for a split-monopole magnetosphere to first order in the small-coupling approximation and second relative order in the slow-rotation approximation. The extracted power at fixed spin and magnetic flux is enhanced in scalar Gauss-Bonnet and reduced in dynamical Chern-Simons gravity, compared to general relativity. We find that there is a degeneracy between spin and the coupling constants of the theories at leading order in the slow rotation approximation that is broken at higher orders.

gr-qc↗