Search

Multi-view diffusion models, obtained by applying Supervised Finetuning (SFT) to text-to-image diffusion models, have driven recent breakthroughs in text-to-3D research. However, due to the limited size and quality of existing 3D datasets, they still suffer from multi-view inconsistencies and Neural Radiance Field (NeRF) reconstruction artifacts. We argue that multi-view diffusion models can benefit from further Reinforcement Learning Finetuning (RLFT), which allows models to learn from the data generated by themselves and improve beyond their dataset limitations during SFT. To this end, we introduce Carve3D, an improved RLFT algorithm coupled with a novel Multi-view Reconstruction Consistency (MRC) metric, to enhance the consistency of multi-view diffusion models. To measure the MRC metric on a set of multi-view images, we compare them with their corresponding NeRF renderings at the same camera viewpoints. The resulting model, which we denote as Carve3DM, demonstrates superior multi-view consistency and NeRF reconstruction quality than existing models. Our results suggest that pairing SFT with Carve3D's RLFT is essential for developing multi-view-consistent diffusion models, mirroring the standard Large Language Model (LLM) alignment pipeline. Our code, training and testing data, and video results are available at: https://desaixie.github.io/carve-3d., Comment: 22 pages, 16 figures. Our code, training and testing data, and video results are available at: https://desaixie.github.io/carve-3d. This paper has been accepted to CVPR 2024. v2: incorporated changes from the CVPR 2024 camera-ready version

We present the first results from a dark matter search using six Skipper-CCDs in the SENSEI detector operating at SNOLAB. With an exposure of 534.9 gram-days from well-performing sensors, we select events containing 2 to 10 electron-hole pairs. After aggressively masking images to remove backgrounds, we observe 55 two-electron events, 4 three-electron events, and no events containing 4 to 10 electrons. The two-electron events are consistent with pileup from one-electron events. Among the 4 three-electron events, 2 appear in pixels that are likely impacted by detector defects, although not strongly enough to trigger our "hot-pixel" mask. We use these data to set world-leading constraints on sub-GeV dark matter interacting with electrons and nuclei., Comment: 5 pages, 2 figures, + Supplemental Materials (5 pages, 5 figures) + References

Context. The determination of accurate photometric redshifts (photo-zs) in large imaging galaxy surveys is key for cosmological studies. One of the most common approaches are machine learning techniques. These methods require a spectroscopic or reference sample to train the algorithms. Attention has to be paid to the quality and properties of these samples since they are key factors in the estimation of reliable photo-zs. Aims. The goal of this work is to calculate the photo-zs for the Y3 DES Deep Fields catalogue using the DNF machine learning algorithm. Moreover, we want to develop techniques to assess the incompleteness of the training sample and metrics to study how incompleteness affects the quality of photometric redshifts. Finally, we are interested in comparing the performance obtained with respect to the EAzY template fitting approach on Y3 DES Deep Fields catalogue. Methods. We have emulated -- at brighter magnitude -- the training incompleteness with a spectroscopic sample whose redshifts are known to have a measurable view of the problem. We have used a principal component analysis to graphically assess incompleteness and to relate it with the performance parameters provided by DNF. Finally, we have applied the results about the incompleteness to the photo-z computation on Y3 DES Deep Fields with DNF and estimated its performance. Results. The photo-zs for the galaxies on DES Deep Fields have been computed with the DNF algorithm and added to the Y3 DES Deep Fields catalogue. They are available at https://des.ncsa.illinois.edu/releases/y3a2/Y3deepfields. Some techniques have been developed to evaluate the performance in the absence of "true" redshift and to assess completeness. We have studied... (Partial abstract), Comment: 14 pages, 17 figures

We present the results of a search for 10--1,000 GeV neutrinos from 2,268 gamma-ray bursts over 8 years of IceCube-DeepCore data. This work probes burst physics below the photosphere where electromagnetic radiation cannot escape. Neutrinos of tens of GeVs are predicted in sub-photospheric collision of free streaming neutrons with bulk-jet protons. In a first analysis, we searched for the most significant neutrino-GRB coincidence using six overlapping time windows centered on the prompt phase of each GRB. In a second analysis, we conducted a search for a group of GRBs, each individually too weak to be detectable, but potentially significant when combined. No evidence of neutrino emission is found for either analysis. The most significant neutrino coincidence is for Fermi-GBM GRB bn 140807500, with a p-value of 0.097 corrected for all trials. The binomial test used to search for a group of GRBs had a p-value of 0.65 after all trial corrections. The binomial test found a group consisting only of GRB bn 140807500 and no additional GRBs. The neutrino limits of this work complement those obtained by IceCube at TeV to PeV energies. We compare our findings for the large set of GRBs as well as GRB 221009A to the sub-photospheric neutron-proton collision model and find that GRB 221009A provides the most constraining limit on baryon loading. For a jet Lorentz factor of 300 (800), the baryon loading on GRB 221009A is lower than 3.85 (2.13) at a 90% confidence level.

Likelihood-free inference is quickly emerging as a powerful tool to perform fast/effective parameter estimation. We demonstrate a technique of optimizing likelihood-free inference to make it even faster by marginalizing symmetries in a physical problem. In this approach, physical symmetries, for example, time-translation are learned using joint-embedding via self-supervised learning with symmetry data augmentations. Subsequently, parameter inference is performed using a normalizing flow where the embedding network is used to summarize the data before conditioning the parameters. We present this approach on two simple physical problems and we show faster convergence in a smaller number of parameters compared to a normalizing flow that does not use a pre-trained symmetry-informed representation., Comment: Accepted for Machine Learning and the Physical Sciences Workshop (submission 69) at NeurIPS 2023; for codes, see https://github.com/ML4GW/summer-projects-2023/blob/neurips-2023/symmetry-informed-flows/README.md

Direct numerical simulations (DNS) are accurate but computationally expensive for predicting materials evolution across timescales, due to the complexity of the underlying evolution equations, the nature of multiscale spatio-temporal interactions, and the need to reach long-time integration. We develop a new method that blends numerical solvers with neural operators to accelerate such simulations. This methodology is based on the integration of a community numerical solver with a U-Net neural operator, enhanced by a temporal-conditioning mechanism that enables accurate extrapolation and efficient time-to-solution predictions of the dynamics. We demonstrate the effectiveness of this framework on simulations of microstructure evolution during physical vapor deposition modeled via the phase-field method. Such simulations exhibit high spatial gradients due to the co-evolution of different material phases with simultaneous slow and fast materials dynamics. We establish accurate extrapolation of the coupled solver with up to 16.5$\times$ speed-up compared to DNS. This methodology is generalizable to a broad range of evolutionary models, from solid mechanics, to fluid dynamics, geophysics, climate, and more.

We present the results of a time-dependent search for neutrino flares in data collected by IceCube between May 2011 and 2021. This data set contains cascade-like events originating from charged-current electron neutrino and tau neutrino interactions and all-flavor neutral-current interactions. IceCube's previous all-sky searches for neutrino flares used data sets consisting of track-like events originating from charged-current muon neutrino interactions. The cascade data sets are statistically independent of the track data sets and provide a new opportunity to observe the transient all-sky landscape. This search uses the spatial, temporal, and energy information of the cascade-like events to conduct searches for the most statistically significant neutrino flares in the northern and southern skies. No statistically significant time-dependent neutrino emission was observed. For the most statistically significant location in the northern sky, $p_\mathrm{global} =$ 0.71, and in the southern sky, $p_\mathrm{global} =$ 0.51. These results are compatible with the background hypothesis. Assuming an E$^{-2.53}$ spectrum from the diffuse astrophysical neutrino flux as measured with cascades, these results are used to calculate upper limits at the 90\% confidence level on neutrino flares of varying duration and constrain the contribution of these flares to the diffuse astrophysical neutrino flux. These constraints are independent of a specified class of astrophysical objects and show that multiple unresolved transient sources may contribute to the diffuse astrophysical neutrino flux., Comment: Submitted to The Astrophysical Journal

This paper explores the use of foundational large language models (LLMs) in hyperparameter optimization (HPO). Hyperparameters are critical in determining the effectiveness of machine learning models, yet their optimization often relies on manual approaches in limited-budget settings. By prompting LLMs with dataset and model descriptions, we develop a methodology where LLMs suggest hyperparameter configurations, which are iteratively refined based on model performance. Our empirical evaluations on standard benchmarks reveal that within constrained search budgets, LLMs can match or outperform traditional HPO methods like Bayesian optimization across different models on standard benchmarks. Furthermore, we propose to treat the code specifying our model as a hyperparameter, which the LLM outputs and affords greater flexibility than existing HPO approaches., Comment: 28 pages

A modern cold triple-axis spectrometer to study quantum condensed matter systems is planned for the High Flux Isotope Reactor (HFIR) at Oak Ridge National Laboratory. Here, we describe the conceptual principles and design of a secondary spectrometer using a multiplexed, prismatic analyzer system relying on graphite crystals and inspired by the successful implementation of the Continuous Angle Multiple Energy Analysis (CAMEA) spectrometers at the Paul Scherrer Institute. This project is currently known as MANTA for Multi-analyzer Neutron Triple-Axis. We report Monte-Carlo ray-tracing simulations on a simple but realistic sample scattering kernel to further illustrate the prismatic analyzer concept's workings, calibration, and performance. Then, we introduce a new statistical analysis approach based on the prismatic analyzer concept to improve the number of final energies measured on the spectrometer. We also study possible evolutions in the CAMEA design relevant for MANTA., Comment: 11 pages, 11 figures

High-precision pulsar timing observations are limited in their accuracy by the jitter noise that appears in the arrival time of pulses. Therefore, it is important to systematically characterise the amplitude of the jitter noise and its variation with frequency. In this paper, we provide jitter measurements from low-frequency wideband observations of PSR J0437$-$4715 using data obtained as part of the Indian Pulsar Timing Array experiment. We were able to detect jitter in both the 300 - 500 MHz and 1260 - 1460 MHz observations of the upgraded Giant Metrewave Radio Telescope (uGMRT). The former is the first jitter measurement for this pulsar below 700 MHz, and the latter is in good agreement with results from previous studies. In addition, at 300 - 500 MHz, we investigated the frequency dependence of the jitter by calculating the jitter for each sub-banded arrival time of pulses. We found that the jitter amplitude increases with frequency. This trend is opposite as compared to previous studies, indicating that there is a turnover at intermediate frequencies. It will be possible to investigate this in more detail with uGMRT observations at 550 - 750 MHz and future high sensitive wideband observations from next generation telescopes, such as the Square Kilometre Array. We also explored the effect of jitter on the high precision dispersion measure (DM) measurements derived from short duration observations. We find that even though the DM precision will be better at lower frequencies due to the smaller amplitude of jitter noise, it will limit the DM precision for high signal-to-noise observations, which are of short durations. This limitation can be overcome by integrating for a long enough duration optimised for a given pulsar., Comment: 13 pages, 12 figures, 3 tables, accepted for Publication of the Astronomical Society of Australia

Low Surface Brightness Galaxies (LSBGs) are excellent probes of quenching and other environmental processes near massive galaxies. We study an extensive sample of LSBGs near massive hosts in the local universe that are distributed across a diverse range of environments. The LSBGs with surface-brightness $\mu_{\rm eff,g}> $24.2 mag arcsec$^{-2}$ are drawn from the Dark Energy Survey Year 3 catalog while the hosts with masses $9.0< log(M_{\star}/M_{\odot})< 11.0$ comparable to the Milky Way and the Large Magellanic Cloud are selected from the z0MGS sample. We study the projected radial density profiles of LSBGs as a function of their color and surface brightness around hosts in both the rich Fornax-Eridanus cluster environment and the low-density field. We detect an overdensity with respect to the background density, out to 2.5 times the virial radius for both hosts in the cluster environment and the isolated field galaxies. When the LSBG sample is split by $g-i$ color or surface brightness $\mu_{\rm eff,g}$, we find the LSBGs closer to their hosts are significantly redder and brighter, like their high surface-brightness counterparts. The LSBGs form a clear 'red sequence' in both the cluster and isolated environments that is visible beyond the virial radius of the hosts. This suggests a pre-processing of infalling LSBGs and a quenched backsplash population around both host samples. However, the relative prominence of the 'blue cloud' feature implies that pre-processing is ongoing near the isolated hosts compared to the cluster hosts., Comment: 22 pages, 7 figures