Search

Formulating an effective constraint model of a parameterised problem class is crucial to the efficiency with which instances of the class can subsequently be solved. It is difficult to know beforehand which of a set of candidate models will perform best in practice. This paper presents a system that employs graph rewriting to reformulate an input model for improved performance automatically. By situating our work in the Essence abstract constraint specification language, we can use the structure in its high level variable types to trigger rewrites directly. We implement our system via rewrite rules expressed in the Graph Programs 2 language, applied to the abstract syntax tree of an input specification. We show how to automatically translate the solution of the reformulated problem into a solution of the original problem for verification and presentation. We demonstrate the efficacy of our system with a detailed case study., Comment: Presented at the PTHG 2024 workshop

We present the Brain Treebank, a large-scale dataset of electrophysiological neural responses, recorded from intracranial probes while 10 subjects watched one or more Hollywood movies. Subjects watched on average 2.6 Hollywood movies, for an average viewing time of 4.3 hours, and a total of 43 hours. The audio track for each movie was transcribed with manual corrections. Word onsets were manually annotated on spectrograms of the audio track for each movie. Each transcript was automatically parsed and manually corrected into the universal dependencies (UD) formalism, assigning a part of speech to every word and a dependency parse to every sentence. In total, subjects heard over 38,000 sentences (223,000 words), while they had on average 168 electrodes implanted. This is the largest dataset of intracranial recordings featuring grounded naturalistic language, one of the largest English UD treebanks in general, and one of only a few UD treebanks aligned to multimodal features. We hope that this dataset serves as a bridge between linguistic concepts, perception, and their neural representations. To that end, we present an analysis of which electrodes are sensitive to language features while also mapping out a rough time course of language processing across these electrodes. The Brain Treebank is available at https://BrainTreebank.dev/, Comment: 36 pages, 17 figures; Accepted at NeurIPS Dataset and Benchmarks 2024

Retrograde analysis is used in game-playing programs to solve states at the end of a game, working backwards toward the start of the game. The algorithm iterates through and computes the perfect-play value for as many states as resources allow. We introduce setrograde analysis which achieves the same results by operating on sets of states that have the same game value. The algorithm is demonstrated by computing exact solutions for Bridge double dummy card-play. For deals with 24 cards remaining to be played ($10^{27}$ states, which can be reduced to $10^{15}$ states using preexisting techniques), we strongly solve all deals. The setrograde algorithm performs a factor of $10^3$ fewer search operations than a standard retrograde algorithm, producing a database with a factor of $10^4$ fewer entries. For applicable domains, this allows retrograde searching to reach unprecedented search depths.

In Partially Observable Markov Decision Processes, integrating an agent's history into memory poses a significant challenge for decision-making. Traditional imitation learning, relying on observation-action pairs for expert demonstrations, fails to capture the expert's memory mechanisms used in decision-making. To capture memory processes as demonstrations, we introduce the concept of memory dependency pairs $(p, q)$ indicating that events at time $p$ are recalled for decision-making at time $q$. We introduce AttentionTuner to leverage memory dependency pairs in Transformers and find significant improvements across several tasks compared to standard Transformers when evaluated on Memory Gym and the Long-term Memory Benchmark. Code is available at https://github.com/WilliamYue37/AttentionTuner.

Superconducting sensors doped with rare isotopes have recently demonstrated powerful sensing performance for sub-keV radiation from nuclear decay. Here, we report the first high-resolution recoil spectroscopy of a single, selected nuclear state using superconducting tunnel junction (STJ) sensors. The STJ sensors were used to measure the eV-scale nuclear recoils produced in $^7$Be electron capture decay in coincidence with the 478 keV $\gamma$-ray emitted in decays to the excited nuclear state in $^7$Li. Details of the Doppler broadened recoil spectrum depend on the slow-down dynamics of the recoil ion and can constrain the interaction potential between the recoiling Li and the Ta matrix of the STJ sensor. The results have implications in several areas from nuclear structure and stopping powers at eV-scale energies to direct searches for dark matter, neutrino mass measurements, and other physics beyond the standard model.

Human excrement composting (HEC) is a sustainable strategy for human excrement (HE) management that recycles nutrients and mitigates health risks while reducing reliance on freshwater, fossil fuels, and fertilizers. We present a comprehensive microbial time series analysis of HEC and show that the initial gut-like microbiome of HEC systems transitions to a microbiome similar to soil and traditional compost in fifteen biological replicates tracked weekly for one year., Comment: Main text: 9 pages, 2 figures; Extended data: 10 figures; Supplemental Text: 32 pages, 8 figures, 2 tables

Crop biomass offers crucial insights into plant health and yield, making it essential for crop science, farming systems, and agricultural research. However, current measurement methods, which are labor-intensive, destructive, and imprecise, hinder large-scale quantification of this trait. To address this limitation, we present a biomass prediction network (BioNet), designed for adaptation across different data modalities, including point clouds and drone imagery. Our BioNet, utilizing a sparse 3D convolutional neural network (CNN) and a transformer-based prediction module, processes point clouds and other 3D data representations to predict biomass. To further extend BioNet for drone imagery, we integrate a neural feature field (NeFF) module, enabling 3D structure reconstruction and the transformation of 2D semantic features from vision foundation models into the corresponding 3D surfaces. For the point cloud modality, BioNet demonstrates superior performance on two public datasets, with an approximate 6.1% relative improvement (RI) over the state-of-the-art. In the RGB image modality, the combination of BioNet and NeFF achieves a 7.9% RI. Additionally, the NeFF-based approach utilizes inexpensive, portable drone-mounted cameras, providing a scalable solution for large field applications.

In autonomous robot navigation, terrain cost assignment is typically performed using a semantics-based paradigm in which terrain is first labeled using a pre-trained semantic classifier and costs are then assigned according to a user-defined mapping between label and cost. While this approach is rapidly adaptable to changing user preferences, only preferences over the types of terrain that are already known by the semantic classifier can be expressed. In this paper, we hypothesize that a machine-learning-based alternative to the semantics-based paradigm above will allow for rapid cost assignment adaptation to preferences expressed over new terrains at deployment time without the need for additional training. To investigate this hypothesis, we introduce and study PACER, a novel approach to costmap generation that accepts as input a single birds-eye view (BEV) image of the surrounding area along with a user-specified preference context and generates a corresponding BEV costmap that aligns with the preference context. Using both real and synthetic data along with a combination of proposed training tasks, we find that PACER is able to adapt quickly to new user preferences while also exhibiting better generalization to novel terrains compared to both semantics-based and representation-learning approaches.

Multimode fibers provide a promising platform for realizing high-power laser amplifiers with suppressed nonlinearities and instabilities. The potential degradation of optical beam quality has been a major concern for highly multimode fiber amplifiers. We show numerically that the beam propagation factor M2 of a single-frequency multimode fiber amplifier can be reduced to nearly unity by shaping the input or output beam profile with spatial phase-masks. Our method works for narrowband multimode fiber amplifiers with strong gain saturation, pump depletion, random mode coupling and polarization mixing. The numerical results validate our approach of utilizing highly multimode excitation to mitigate nonlinear effects in high-power fiber amplifiers and performing input wavefront shaping to control output beam profile and polarization state.

Many nucleosynthetic channels create the elements, but two-parameter models characterized by $\alpha$ and Fe nonetheless predict stellar abundances in the Galactic disk to accuracies of 0.02 to 0.05 dex for most measured elements, near the level of current abundance uncertainties. It is difficult to make individual measurements more precise than this to investigate lower-amplitude nucleosynthetic effects, but population studies of mean abundance patterns can reveal more subtle abundance differences. Here we look at the detailed abundances for 67315 stars from APOGEE DR17, but in the form of abundance residuals away from a best-fit two-parameter, data-driven nucleosynthetic model. We find that these residuals show complex structures with respect to age, guiding radius, and vertical action that are not random and are also not strongly correlated with sources of systematic error such as surface gravity, effective temperature, and radial velocity. The residual patterns, especially in Na, C+N, Ni, Mn, and Ce, trace kinematic structures in the Milky Way, such as the inner disk, thick disk, and flared outer disk. A principal component analysis suggests that most of the observed structure is low-dimensional and can be explained by a few eigenvectors. We find that some, but not all, of the effects in the low-$\alpha$ disk can be explained by dilution with fresh gas, so that abundance ratios resemble those of stars with higher metallicity. The patterns and maps we provide could be combined with accurate forward models of nucleosynthesis, star formation, and gas infall to provide a more detailed picture of star and element formation in different Milky Way components.

Vascular diseases such as thrombosis, atherosclerosis, and aneurysm, which can lead to blockage of blood flow or blood vessel rupture, are common and life-threatening. Conventional minimally invasive treatments utilize catheters, or long tubes, to guide small devices or therapeutic agents to targeted regions for intervention. Unfortunately, catheters suffer from difficult and unreliable navigation in narrow, winding vessels such as those found in the brain. Magnetically actuated untethered robots, which have been extensively explored as an alternative, are promising for navigation in complex vasculatures and vascular disease treatments. Most current robots, however, cannot swim against high flows or are inadequate in treating certain conditions. Here, we introduce a multifunctional and magnetically actuated milli-spinner robot for rapid navigation and performance of various treatments in complicated vasculatures. The milli-spinner, with a unique hollow structure including helical fins and slits for propulsion, generates a distinct flow field upon spinning. The milli-spinner is the fastest-ever untethered magnetic robot for movement in tubular environments, easily achieving speeds of 23 cm/s, demonstrating promise as an untethered medical device for effective navigation in blood vessels and robotic treatment of numerous vascular diseases.

Galaxy Zoo is an online project to classify morphological features in extra-galactic imaging surveys with public voting. In this paper, we compare the classifications made for two different surveys, the Dark Energy Spectroscopic Instrument (DESI) imaging survey and a part of the Kilo-Degree Survey (KiDS), in the equatorial fields of the Galaxy And Mass Assembly (GAMA) survey. Our aim is to cross-validate and compare the classifications based on different imaging quality and depth. We find that generally the voting agrees globally but with substantial scatter i.e. substantial differences for individual galaxies. There is a notable higher voting fraction in favor of ``smooth'' galaxies in the DESI+\rev{{\sc zoobot}} classifications, most likely due to the difference between imaging depth. DESI imaging is shallower and slightly lower resolution than KiDS and the Galaxy Zoo images do not reveal details such as disk features \rev{and thus are missed in the {\sc zoobot} training sample}. \rev{We check against expert visual classifications and find good agreement with KiDS-based Galaxy Zoo voting.} We reproduce the results from Porter-Temple+ (2022), on the dependence of stellar mass, star-formation, and specific star-formation on the number of spiral arms. This shows that once corrected for redshift, the DESI Galaxy Zoo and KiDS Galaxy Zoo classifications agree well on population properties. The zoobot cross-validation increases confidence in its ability to compliment Galaxy Zoo classifications and its ability for transfer learning across surveys., Comment: 20 pages, 22 figures, 8 tables, accepted for publication in PASA

Stellar-mass black holes (BHs) embedded in active galactic nuclei (AGN) may be major sources of astrophysical gravitational waves (GWs), contributing both to the observed LIGO-Virgo-KAGRA population of binary BH mergers and to future populations of LISA-band extreme mass ratio inspirals (EMRIs). The ability of these BHs to pair up into binaries, inspiral, and produce GWs will be shaped by the existence of migration traps, regions in the AGN where hydrodynamic torques vanish. Previous works have studied the existence and location of migration traps in AGN disks. Here, we investigate how individual BHs may escape such traps as an outcome of mergers, potentially suppressing hierarchical growth. We find that while GW recoil kicks are strong enough to kick merged BHs onto inclined orbits, gas drag quickly realigns them into the AGN disk. A more robust escape mechanism is gap opening: once a BH grows above a critical mass, its gravity disturbs the AGN gas sufficiently to eliminate the trap. In low-mass AGN relevant for LISA, gaps open easily and the resulting ``wet EMRI'' masses are unlikely to reflect protracted hierarchical mergers. In combination with our previous work, we find that migration traps only exist in a relatively narrow range of AGN luminosities between [10^{43.5},10^{45.5}] erg/s. We identify an even narrower AGN luminosity range for which stellar mass BHs can grow into the pair instability mass gap and beyond. This characteristic luminosity scale may assist in indirect tests of the ``AGN channel'' for binary BH mergers., Comment: "8 pages, 4 figures in main text. Comments welcome!

Unsupervised skill discovery carries the promise that an intelligent agent can learn reusable skills through autonomous, reward-free environment interaction. Existing unsupervised skill discovery methods learn skills by encouraging distinguishable behaviors that cover diverse states. However, in complex environments with many state factors (e.g., household environments with many objects), learning skills that cover all possible states is impossible, and naively encouraging state diversity often leads to simple skills that are not ideal for solving downstream tasks. This work introduces Skill Discovery from Local Dependencies (Skild), which leverages state factorization as a natural inductive bias to guide the skill learning process. The key intuition guiding Skild is that skills that induce diverse interactions between state factors are often more valuable for solving downstream tasks. To this end, Skild develops a novel skill learning objective that explicitly encourages the mastering of skills that effectively induce different interactions within an environment. We evaluate Skild in several domains with challenging, long-horizon sparse reward tasks including a realistic simulated household robot domain, where Skild successfully learns skills with clear semantic meaning and shows superior performance compared to existing unsupervised reinforcement learning methods that only maximize state coverage.

Many robot manipulation tasks require active or interactive exploration behavior in order to be performed successfully. Such tasks are ubiquitous in embodied domains, where agents must actively search for the information necessary for each stage of a task, e.g., moving the head of the robot to find information relevant to manipulation, or in multi-robot domains, where one scout robot may search for the information that another robot needs to make informed decisions. We identify these tasks with a new type of problem, factorized Contextual Markov Decision Processes, and propose DISaM, a dual-policy solution composed of an information-seeking policy that explores the environment to find the relevant contextual information and an information-receiving policy that exploits the context to achieve the manipulation goal. This factorization allows us to train both policies separately, using the information-receiving one to provide reward to train the information-seeking policy. At test time, the dual agent balances exploration and exploitation based on the uncertainty the manipulation policy has on what the next best action is. We demonstrate the capabilities of our dual policy solution in five manipulation tasks that require information-seeking behaviors, both in simulation and in the real-world, where DISaM significantly outperforms existing methods. More information at https://robin-lab.cs.utexas.edu/learning2look/., Comment: Project Website: https://robin-lab.cs.utexas.edu/learning2look/

We report a surface instability observed during the extrusion of extremely soft elastic solids in confined geometries. Due to their unique rheological properties, these soft solids can migrate through narrow gaps by continuously everting the bulk material. The extrusion front spontaneously buckles in the direction transverse to the flow, resulting in a furrow-like morphology that deepens over time. We characterize the distinct features of this instability using experiments and theory and contrast the results with known elastic surface instabilities. Our study may provide insights into various processes involving the extrusion-like deformations of soft solids, including biomaterials.

Massive black hole (MBH) binaries in galactic nuclei are one of the leading sources of $\sim$ mHz gravitational waves (GWs) for future missions such as $\rm{\textit{LISA}}$. However, the poor sky localization of GW interferometers will make it challenging to identify the host galaxy of MBH mergers absent an electromagnetic counterpart. One such counterpart is the tidal disruption of a star that has been captured into mean motion resonance with the inspiraling binary. Here we investigate the production of tidal disruption events (TDEs) through capture into, and subsequent evolution in, orbital resonance. We examine the full nonlinear evolution of planar autoresonance for stars that lock in to autoresonance with a shrinking MBH binary. Capture into the 2:1 resonance is guaranteed for any realistic astrophysical parameters (given a relatively small MBH binary mass ratio), and the captured star eventually attains an eccentricity $e\approx 1$, leading to a TDE. Stellar disks can be produced around MBHs following an active galactic nucleus episode, and we estimate the TDE rates from resonant capture produced when a secondary MBH begins inspiralling through such a disk. In some cases, the last resonant TDE can occur within a decade of the eventual $\rm{\textit{LISA}}$ signal, helping to localize the GW event.

A hallmark of intelligent agents is the ability to learn reusable skills purely from unsupervised interaction with the environment. However, existing unsupervised skill discovery methods often learn entangled skills where one skill variable simultaneously influences many entities in the environment, making downstream skill chaining extremely challenging. We propose Disentangled Unsupervised Skill Discovery (DUSDi), a method for learning disentangled skills that can be efficiently reused to solve downstream tasks. DUSDi decomposes skills into disentangled components, where each skill component only affects one factor of the state space. Importantly, these skill components can be concurrently composed to generate low-level actions, and efficiently chained to tackle downstream tasks through hierarchical Reinforcement Learning. DUSDi defines a novel mutual-information-based objective to enforce disentanglement between the influences of different skill components, and utilizes value factorization to optimize this objective efficiently. Evaluated in a set of challenging environments, DUSDi successfully learns disentangled skills, and significantly outperforms previous skill discovery methods when it comes to applying the learned skills to solve downstream tasks. Code and skills visualization at jiahenghu.github.io/DUSDi-site/., Comment: NeurIPS2024

Fragmentation in a gravitationally unstable accretion disk can be an important pathway for forming stellar/planetary companions. To characterize quantitatively the condition for and outcome of fragmentation under realistic thermodynamics, we perform global 3D simulations of gravitationally unstable disks at various cooling rates and cooling types, including the first global simulations of gravitational instability that employ full radiation transport. We find that fragmentation is a stochastic process, with the fragment generation rate per disk area $p_{\rm frag}$ showing an exponential dependence on the parameter $\beta\equiv\Omega_K t_{cool}$, where $\Omega_K$ is the Keplerian rotation frequency and $t_{cool}$ is the average cooling timescale. Compared to a prescribed constant $\beta$, radiative cooling in the optically thin/thick regime makes $p_{\rm frag}$ decrease slower/faster in $\beta$; the critical $\beta$ corresponding to $\sim 1$ fragment per orbit is $\approx$3, 5, 2 for constant $\beta$, optically thin, and optically thick cooling, respectively. The distribution function of the initial fragment mass is remarkably insensitive to disk thermodynamics. Regardless of cooling rate and optical depth, the typical initial fragment mass is $m_{frag} \approx 40 M_{tot}h^3$, with $M_{tot}$ being the total (star+disk) mass and $h=H/R$ being the disk aspect ratio. Applying this result to typical Class 0/I protostellar disks, we find $m_{frag}\sim 20 M_J$, suggesting that fragmentation more likely forms brown dwarfs. Given the finite width of the $m_{frag}$ distribution, forming massive planets is also possible., Comment: 21 pages, 11 figures, accepted for publication in ApJ

Light passing near a black hole can follow multiple paths from an emission source to an observer due to strong gravitational lensing. Photons following different paths take different amounts of time to reach the observer, which produces an echo signature in the image. The characteristic echo delay is determined primarily by the mass of the black hole, but it is also influenced by the black hole spin and inclination to the observer. In the Kerr geometry, echo images are demagnified, rotated, and sheared copies of the direct image and lie within a restricted region of the image. Echo images have exponentially suppressed flux, and temporal correlations within the flow make it challenging to directly detect light echoes from the total light curve. In this paper, we propose a novel method to search for light echoes by correlating the total light curve with the interferometric signal at high spatial frequencies, which is a proxy for indirect emission. We explore the viability of our method using numerical general relativistic magnetohydrodynamic simulations of a near-face-on accretion system scaled to M87-like parameters. We demonstrate that our method can be used to directly infer the echo delay period in simulated data. An echo detection would be clear evidence that we have captured photons that have circled the black hole, and a high-fidelity echo measurement would provide an independent measure of fundamental black hole parameters. Our results suggest that detecting echoes may be achievable through interferometric observations with a modest space-based very long baseline interferometry mission., Comment: 21 pages, 15 figures, accepted for publication in ApJL

Recent advances in CV and NLP have been largely driven by scaling up the number of network parameters, despite traditional theories suggesting that larger networks are prone to overfitting. These large networks avoid overfitting by integrating components that induce a simplicity bias, guiding models toward simple and generalizable solutions. However, in deep RL, designing and scaling up networks have been less explored. Motivated by this opportunity, we present SimBa, an architecture designed to scale up parameters in deep RL by injecting a simplicity bias. SimBa consists of three components: (i) an observation normalization layer that standardizes inputs with running statistics, (ii) a residual feedforward block to provide a linear pathway from the input to output, and (iii) a layer normalization to control feature magnitudes. By scaling up parameters with SimBa, the sample efficiency of various deep RL algorithms-including off-policy, on-policy, and unsupervised methods-is consistently improved. Moreover, solely by integrating SimBa architecture into SAC, it matches or surpasses state-of-the-art deep RL methods with high computational efficiency across DMC, MyoSuite, and HumanoidBench. These results demonstrate SimBa's broad applicability and effectiveness across diverse RL algorithms and environments., Comment: preprint

In robot planning, tasks can often be achieved through multiple options, each consisting of several actions. This work specifically addresses deadline constraints in task and motion planning, aiming to find a plan that can be executed within the deadline despite uncertain planning and execution times. We propose an effort allocation problem, formulated as a Markov decision process (MDP), to find such a plan by leveraging metareasoning perspectives to allocate computational resources among the given options. We formally prove the NP-hardness of the problem by reducing it from the knapsack problem. Both a model-based approach, where transition models are learned from past experience, and a model-free approach, which overcomes the unavailability of prior data acquisition through reinforcement learning, are explored. For the model-based approach, we investigate Monte Carlo tree search (MCTS) to approximately solve the proposed MDP and further design heuristic schemes to tackle NP-hardness, leading to the approximate yet efficient algorithm called DP_Rerun. In experiments, DP_Rerun demonstrates promising performance comparable to MCTS while requiring negligible computation time., Comment: 48 pages, 6 figures

In order to train agents that can quickly adapt to new objectives or reward functions, efficient unsupervised representation learning in sequential decision-making environments can be important. Frameworks such as the Exogenous Block Markov Decision Process (Ex-BMDP) have been proposed to formalize this representation-learning problem (Efroni et al., 2022b). In the Ex-BMDP framework, the agent's high-dimensional observations of the environment have two latent factors: a controllable factor, which evolves deterministically within a small state space according to the agent's actions, and an exogenous factor, which represents time-correlated noise, and can be highly complex. The goal of the representation learning problem is to learn an encoder that maps from observations into the controllable latent space, as well as the dynamics of this space. Efroni et al. (2022b) has shown that this is possible with a sample complexity that depends only on the size of the controllable latent space, and not on the size of the noise factor. However, this prior work has focused on the episodic setting, where the controllable latent state resets to a specific start state after a finite horizon. By contrast, if the agent can only interact with the environment in a single continuous trajectory, prior works have not established sample-complexity bounds. We propose STEEL, the first provably sample-efficient algorithm for learning the controllable dynamics of an Ex-BMDP from a single trajectory, in the function approximation setting. STEEL has a sample complexity that depends only on the sizes of the controllable latent space and the encoder function class, and (at worst linearly) on the mixing time of the exogenous noise factor. We prove that STEEL is correct and sample-efficient, and demonstrate STEEL on two toy problems. Code is available at: https://github.com/midi-lab/steel.

Simulation has enabled unprecedented compute-scalable approaches to robot learning. However, many existing simulation frameworks typically support a narrow range of scenes/tasks and lack features critical for scaling generalizable robotics and sim2real. We introduce and open source ManiSkill3, the fastest state-visual GPU parallelized robotics simulator with contact-rich physics targeting generalizable manipulation. ManiSkill3 supports GPU parallelization of many aspects including simulation+rendering, heterogeneous simulation, pointclouds/voxels visual input, and more. Simulation with rendering on ManiSkill3 can run 10-1000x faster with 2-3x less GPU memory usage than other platforms, achieving up to 30,000+ FPS in benchmarked environments due to minimal python/pytorch overhead in the system, simulation on the GPU, and the use of the SAPIEN parallel rendering system. Tasks that used to take hours to train can now take minutes. We further provide the most comprehensive range of GPU parallelized environments/tasks spanning 12 distinct domains including but not limited to mobile manipulation for tasks such as drawing, humanoids, and dextrous manipulation in realistic scenes designed by artists or real-world digital twins. In addition, millions of demonstration frames are provided from motion planning, RL, and teleoperation. ManiSkill3 also provides a comprehensive set of baselines that span popular RL and learning-from-demonstrations algorithms., Comment: Project website: http://maniskill.ai/

A fundamental challenge in continual learning is to balance the trade-off between learning new tasks and remembering the previously acquired knowledge. Gradient Episodic Memory (GEM) achieves this balance by utilizing a subset of past training samples to restrict the update direction of the model parameters. In this work, we start by analyzing an often overlooked hyper-parameter in GEM, the memory strength, which boosts the empirical performance by further constraining the update direction. We show that memory strength is effective mainly because it improves GEM's generalization ability and therefore leads to a more favorable trade-off. By this finding, we propose two approaches that more flexibly constrain the update direction. Our methods are able to achieve uniformly better Pareto Frontiers of remembering old and learning new knowledge than using memory strength. We further propose a computationally efficient method to approximately solve the optimization problem with more constraints.

The high cost of real-world data for robotics Reinforcement Learning (RL) leads to the wide usage of simulators. Despite extensive work on building better dynamics models for simulators to match with the real world, there is another, often-overlooked mismatch between simulations and the real world, namely the distribution of available training tasks. Such a mismatch is further exacerbated by existing curriculum learning techniques, which automatically vary the simulation task distribution without considering its relevance to the real world. Considering these challenges, we posit that curriculum learning for robotics RL needs to be grounded in real-world task distributions. To this end, we propose Grounded Curriculum Learning (GCL), which aligns the simulated task distribution in the curriculum with the real world, as well as explicitly considers what tasks have been given to the robot and how the robot has performed in the past. We validate GCL using the BARN dataset on complex navigation tasks, achieving a 6.8% and 6.5% higher success rate compared to a state-of-the-art CL method and a curriculum designed by human experts, respectively. These results show that GCL can enhance learning efficiency and navigation performance by grounding the simulation task distribution in the real world within an adaptive curriculum., Comment: 8 pages, 4 figures

A search for proton decay into $e^+/\mu^+$ and a $\eta$ meson has been performed using data from a 0.373 Mton$\cdot$year exposure (6050.3 live days) of Super-Kamiokande. Compared to previous searches this work introduces an improved model of the intranuclear $\eta$ interaction cross section, resulting in a factor of two reduction in uncertainties from this source and $\sim$10\% increase in signal efficiency. No significant data excess was found above the expected number of atmospheric neutrino background events resulting in no indication of proton decay into either mode. Lower limits on the proton partial lifetime of $1.4\times\mathrm{10^{34}~years}$ for $p\rightarrow e^+\eta$ and $7.3\times\mathrm{10^{33}~years}$ for $p\rightarrow \mu^+\eta$ at the 90$\%$ C.L. were set. These limits are around 1.5 times longer than our previous study and are the most stringent to date.

The Beryllium Electron capture in Superconducting Tunnel junctions (BeEST) experiment searches for evidence of heavy neutrino mass eigenstates in the nuclear electron capture decay of $^7$Be by precisely measuring the recoil energy of the $^7$Li daughter. In Phase-III, the BeEST experiment has been scaled from a single superconducting tunnel junction (STJ) sensor to a 36-pixel array to increase sensitivity and mitigate gamma-induced backgrounds. Phase-III also uses a new continuous data acquisition system that greatly increases the flexibility for signal processing and data cleaning. We have developed procedures for signal processing and spectral fitting that are sufficiently robust to be automated for large data sets. This article presents the optimized procedures before unblinding the majority of the Phase-III data set to search for physics beyond the standard model.

The scenario of feedback-free starbursts (FFB), which predicts excessively bright galaxies at cosmic dawn as observed using JWST, may provide a natural setting for black hole (BH) growth. This involves the formation of intermediate-mass seed BHs and their runaway mergers into super-massive BHs with high BH-to-stellar mass ratios and low AGN luminosities. We present a scenario of merger-driven BH growth in FFB galaxies and study its feasibility. BH seeds form within the building blocks of the FFB galaxies, namely, thousands of compact star clusters, each starbursting in a free-fall time of a few Myr before the onset of stellar and supernova feedback. The BH seeds form by rapid core collapse in the FFB clusters, in a few free-fall times, sped up by the migration of massive stars due to the young, broad stellar mass function and stimulated by a `gravo-gyro' instability due to internal cluster rotation and flattening. BHs of $10^4 M_\odot$ are expected in $10^6 M_\odot$ FFB clusters within sub-kpc galactic disks at $z \sim 10$. The BHs then migrate to the galaxy center by dynamical friction, hastened by the compact FFB stellar galactic disk configuration. Efficient mergers of the BH seeds will produce $10^{6-8} M_\odot$ BHs with a BH-to-stellar mass ratio $\sim 0.01$ by $z \sim 4-7$, as observed. The growth of the central BH by mergers can overcome the bottleneck introduced by gravitational wave recoils if the BHs inspiral within a relatively cold disk or if the escape velocity from the galaxy is boosted by a wet compaction event. Such events, common in massive galaxies at high redshifts, can also help by speeding up the inward BH migration and by providing central gas to assist with the final parsec problem. The cold disk version of the FFB scenario provides a feasible route for the formation of supermassive BHs., Comment: 24 pages, 14 figures

TOI-1266 is a benchmark system of two temperate ($<$ 450 K) sub-Neptune-sized planets orbiting a nearby M dwarf exhibiting a rare inverted architecture with a larger interior planet. In this study, we characterize transit timing variations (TTVs) in the TOI-1266 system using high-precision ground-based follow-up and new TESS data. We confirm the presence of a third exterior non-transiting planet, TOI-1266 d (P = 32.5 d, $M_d$ = 3.68$^{+1.05}_{-1.11} M_{\oplus}$), and combine the TTVs with archival radial velocity (RV) measurements to improve our knowledge of the planetary masses and radii. We find that, consistent with previous studies, TOI-1266 b ($R_b$ = 2.52 $\pm$ 0.08 $R_{\oplus}$, $M_b$ = 4.46 $\pm$ 0.69 $M_{\oplus}$) has a low bulk density requiring the presence of a hydrogen-rich envelope, while TOI-1266 c ($R_c$ = 1.98 $\pm$ 0.10 $R_{\oplus}$, $M_c$ = 3.17 $\pm$ 0.76 $M_{\oplus}$) has a higher bulk density that can be matched by either a hydrogen-rich or water-rich envelope. Our new dynamical model reveals that this system is arranged in a rare configuration with the inner and outer planets located near the 3:1 period ratio with a non-resonant planet in between them. Our dynamical fits indicate that the inner and outer planet have significantly nonzero eccentricities ($e_b + e_d = 0.076^{+0.029}_{-0.019}$), suggesting that TOI-1266 b may have an inflated envelope due to tidal heating. Finally, we explore the corresponding implications for the formation and long-term evolution of the system, which contains two of the most favorable cool ($<$ 500 K) sub-Neptunes for atmospheric characterization with JWST., Comment: 36 pages, 19 figures, 5 tables, submitted to The Astronomical Journal

We describe AthenaK: a new implementation of the Athena++ block-based adaptive mesh refinement (AMR) framework using the Kokkos programming model. Finite volume methods for Newtonian, special relativistic (SR), and general relativistic (GR) hydrodynamics and magnetohydrodynamics (MHD), and GR-radiation hydrodynamics and MHD, as well as a module for evolving Lagrangian tracer or charged test particles (e.g., cosmic rays) are implemented using the framework. In two companion papers we describe (1) a new solver for the Einstein equations based on the Z4c formalism and (2) a GRMHD solver in dynamical spacetimes also implemented using the framework, enabling new applications in numerical relativity. By adopting Kokkos, the code can be run on virtually any hardware, including CPUs, GPUs from multiple vendors, and emerging ARM processors. AthenaK shows excellent performance and weak scaling, achieving over one billion cell updates per second for hydrodynamics in three-dimensions on a single NVIDIA Grace Hopper processor and with a typical parallel efficiency of 80% on 65536 AMD GPUs on the OLCF Frontier system. Such performance portability enables AthenaK to leverage modern exascale computing systems for challenging applications in astrophysical fluid dynamics, numerical relativity, and multimessenger astrophysics., Comment: Submitted to ApJS

We introduce a learning-guided motion planning framework that provides initial seed trajectories using a diffusion model for trajectory optimization. Given a workspace, our method approximates the configuration space (C-space) obstacles through a key-configuration representation that consists of a sparse set of task-related key configurations, and uses this as an input to the diffusion model. The diffusion model integrates regularization terms that encourage collision avoidance and smooth trajectories during training, and trajectory optimization refines the generated seed trajectories to further correct any colliding segments. Our experimental results demonstrate that using high-quality trajectory priors, learned through our C-space-grounded diffusion model, enables efficient generation of collision-free trajectories in narrow-passage environments, outperforming prior learning- and planning-based baselines. Videos and additional materials can be found on the project page: https://kiwi-sherbet.github.io/PRESTO., Comment: Submitted to ICRA 2025

In recent years, the Robotics field has initiated several efforts toward building generalist robot policies through large-scale multi-task Behavior Cloning. However, direct deployments of these policies have led to unsatisfactory performance, where the policy struggles with unseen states and tasks. How can we break through the performance plateau of these models and elevate their capabilities to new heights? In this paper, we propose FLaRe, a large-scale Reinforcement Learning fine-tuning framework that integrates robust pre-trained representations, large-scale training, and gradient stabilization techniques. Our method aligns pre-trained policies towards task completion, achieving state-of-the-art (SoTA) performance both on previously demonstrated and on entirely novel tasks and embodiments. Specifically, on a set of long-horizon mobile manipulation tasks, FLaRe achieves an average success rate of 79.5% in unseen environments, with absolute improvements of +23.6% in simulation and +30.7% on real robots over prior SoTA methods. By utilizing only sparse rewards, our approach can enable generalizing to new capabilities beyond the pretraining data with minimal human effort. Moreover, we demonstrate rapid adaptation to new embodiments and behaviors with less than a day of fine-tuning. Videos can be found on the project website at https://robot-flare.github.io/

We report the fabrication and characterisation of a multi-core anti-resonant hollow core fibre with low inter-core coupling. The optical losses were 0.03 and 0.08 dB/m at 620 and 1000 nm respectively, while the novel structure provides new insights into hollow core fibre design and fabrication.

We present a velocity-resolved reverberation mapping analysis of the hypervariable quasar RM160 (SDSS J141041.25+531849.0) at z = 0.359 with 153 spectroscopic epochs of data representing a ten-year baseline (2013-2023). We split the baseline into two regimes based on the 3x flux increase in the light curve: a 'low state' phase during the years 2013-2019 and a 'high state' phase during the years 2022-2023. The velocity-resolved lag profiles (VRLP) indicate that gas with different kinematics dominates the line emission in different states. The H\b{eta} VRLP begins with a signature of inflow onto the BLR in the 'low state', while in the 'high state' it is flatter with less signature of inflow. The H{\alpha} VRLP begins consistent with a virialized BLR in the 'low state', while in the 'high state' shows a signature of inflow. The differences in the kinematics between the Balmer lines and between the 'low state' and the 'high state' suggests complex BLR dynamics. We find that the BLR radius and velocity (both FWHM and {\sigma}) do not obey a constant virial product throughout the monitoring period. We find that BLR lags and continuum luminosity are correlated, consistent with rapid response of the BLR gas to the illuminating continuum. The BLR kinematic profile changes in unpredictable ways that are not related to continuum changes and reverberation lag. Our observations indicate that non-virial kinematics can significantly contribute to observed line profiles, suggesting caution for black-hole mass estimation in luminous and highly varying quasars like RM160., Comment: 23 pages, 17 figures

We introduce an extension to the AthenaK code for general-relativistic magnetohydrodynamics (GRMHD) in dynamical spacetimes using a 3+1 conservative Eulerian formulation. Like the fixed-spacetime GRMHD solver, we use standard finite-volume methods to evolve the fluid and a constrained transport scheme to preserve the divergence-free constraint for the magnetic field. We also utilize a first-order flux correction (FOFC) scheme to reduce the need for an artificial atmosphere and optionally enforce a maximum principle to improve robustness. We demonstrate the accuracy of AthenaK using a set of standard tests in flat and curved spacetimes. Using a SANE accretion disk around a Kerr black hole, we compare the new solver to the existing solver for stationary spacetimes using the so-called "HARM-like" formulation. We find that both formulations converge to similar results. We also include the first published binary neutron star (BNS) mergers performed on graphical processing units (GPUs). Thanks to the FOFC scheme, our BNS mergers maintain a relative error of $\mathcal{O}(10^{-11})$ or better in baryon mass conservation up to collapse. Finally, we perform scaling tests of AthenaK on OLCF Frontier, where we show excellent weak scaling of $\geq 80\%$ efficiency up to 32768 GPUs and $74\%$ up to 65536 GPUs for a GRMHD problem in dynamical spacetimes with six levels of mesh refinement. AthenaK achieves an order-of-magnitude speedup using GPUs compared to CPUs, demonstrating that it is suitable for performing numerical relativity problems on modern exascale resources., Comment: 26 pages, 25 figures; to be submitted to ApJS soon

We present the numerical relativity module within AthenaK, an open source performance-portable astrophysics code designed for exascale computing applications. This module employs the Z4c formulation to solve the Einstein equations. We demonstrate its accuracy through a series of standard numerical relativity tests, including convergence of the gravitational waveform from binary black hole coalescence. Furthermore, we conduct scaling tests on OLCF Frontier and NERSC Perlmutter, where AthenaK exhibits excellent weak scaling efficiency of 80% on up to 65,536 AMD MI250X GPUs on Frontier (relative to 4 GPUs) and strong scaling efficiencies of 84% and 77% on AMD MI250X and NVIDIA A100 GPUs on Frontier and Perlmutter respectively. Additionally, we observe a significant performance boost, with two orders of magnitude speedup ($\gtrsim 200\times$) on a GPU compared to a single CPU core, affirming that AthenaK is well-suited for exascale computing, thereby expanding the potential for breakthroughs in numerical relativity research.

Polymer mixtures fractionate between phases depending on their molecular weight. Consequently, by varying solvent conditions, a polydisperse polymer sample can be separated between phases so as to achieve a particular molecular weight distribution in each phase. In principle, predictive physics-based theories can help guide separation design and interpret experimental fractionation measurements. Even so, applying the standard Flory-Huggins model can present a computational challenge for mixtures with many polymeric components of different length, particularly for scarce components at the tails of a distribution. Here, we apply our recently-derived exact analytical solution of multi-component Flory-Huggins theory for polydisperse polymers to understand the principles of polymer fractionation for common molecular weight distributions. Our method reveals that polymer fractionation is highly sensitive to the shape, and in particular the tails, of this distribution. Our results highlight the need for considering the full molecular weight distribution in phase coexistence calculations.

We analyze the steady viscoelastic fluid flow in slowly varying contracting channels of arbitrary shape and present a theory based on the lubrication approximation for calculating the flow rate-pressure drop relation at low and high Deborah ($De$) numbers. Unlike most prior theoretical studies leveraging the Oldroyd-B model, we describe the fluid viscoelasticity using a FENE-CR model and examine how the polymer chains' finite extensibility impacts the pressure drop. We employ the low-Deborah-number lubrication analysis to provide analytical expressions for the pressure drop up to $O(De^4)$. We further consider the ultra-dilute limit and exploit a one-way coupling between the parabolic velocity and elastic stresses to calculate the pressure drop of the FENE-CR fluid for arbitrary values of the Deborah number. Such an approach allows us to elucidate elastic stress contributions governing the pressure drop variations and the effect of finite extensibility for all $De$. We validate our theoretical predictions with two-dimensional numerical simulations and find excellent agreement. We show that, at low Deborah numbers, the pressure drop of the FENE-CR fluid monotonically decreases with $De$, similar to the previous results for the Oldroyd-B and FENE-P fluids. However, at high Deborah numbers, in contrast to a linear decrease for the Oldroyd-B fluid, the pressure drop of the FENE-CR fluid exhibits a non-monotonic variation due to finite extensibility, first decreasing and then increasing with $De$. Nevertheless, even at sufficiently high Deborah numbers, the pressure drop of the FENE-CR fluid in the ultra-dilute and lubrication limits is lower than the corresponding Newtonian pressure drop.

Over the past two decades, tight correlations between black hole masses ($M_\bullet$) and their host galaxy properties have been firmly established for massive galaxies ($\log(M_*/M_{\odot})\gtrsim10$) at low-$z$ ($z<1$), indicating coevolution of supermassive black holes and galaxies. However, the situation at high-$z$, especially beyond cosmic noon ($z\gtrsim2.5$), is controversial. With a combination of \emph{JWST} NIRCam/wide field slitless spectroscopy (WFSS) from FRESCO, CONGRESS and deep multi-band NIRCam/image data from JADES in the GOODS fields, we study the black hole to galaxy mass relation at z$\sim$1--4. After identifying 18 broad-line active galactic nuclei (BL AGNs) at $12.5$) from the WFSS data, we measure their black hole masses based on broad near-infrared lines (Pa $\alpha$, Pa $\beta$, and He\,I $\lambda$10833\,\AA), and constrain their stellar masses ($M_{*}$) from AGN-galaxy image decomposition or SED decomposition. Taking account of the observational biases, the intrinsic scatter of the $M_{\bullet}-M_{*}$ relation, and the errors in mass measurements, we find no significant difference in the $M_{\bullet}/M_{*}$ ratio for 2.5 $< $ z $ <$ 4 compared to that at lower redshifts ($1 < z < 2.5$), suggesting no evolution of the $M_{\bullet} - M_{*}$ relation at $\log(M_*/M_{\odot})\gtrsim10$ up to z$\sim$4., Comment: 23 pages, 12 figures, submitted to AAS Journals

Highly inclined (edge-on) disk galaxies offer the unique perspective to constrain their intrinsic flattening, $c/a$, where $c$ and $a$ are respectively the vertical and long radial axes of the disk measured at suitable stellar densities. The ratio $c/a$ is a necessary quantity in the assessment of galaxy inclinations, three-dimensional structural reconstructions, total masses, as well as a constraint to galaxy formation models. 3.6 micron maps of 133 edge-on spiral galaxies from the Spitzer Survey of Stellar Structure in Galaxies (S4G) and its early-type galaxy extension are used to revisit the assessment of $c/a$ free from dust extinction and away from the influence of a stellar bulge. We present a simple definition of $c/a$ and explore trends with other galactic physical parameters: total stellar mass, concentration index, total HI mass, mass of the central mass concentration, circular velocity, model-dependent scales, as well as Hubble type. Other than a dependence on early/late Hubble types, and a related trend with light concentration, no other parameters were found to correlate with the intrinsic flattening of spiral galaxies. The latter is mostly constant with $\langle c/a \rangle$ = 0.124 $\pm$ 0.001 (stat) $\pm$ 0.033 (intrinsic/systematic) and greater for earlier types., Comment: 27 full pages, 14 figures, 6 tables, 1 appendix. Submitted to AAS Journals. Comments welcome

In addressing the Department of Energy's April, 2022 announcement of a Bold Decadal Vision for delivering a Fusion Pilot Plant by 2035, associated software tools need to be developed for the integration of real world engineering and supply chain data with advanced science models that are accelerated with Machine Learning. An associated research and development effort has been introduced here with promising early progress on the delivery of a realistic Digital Twin Tokamak that has benefited from accelerated advances by the Princeton University AI Deep Learning innovative near-real-time simulators accompanied by technological capabilities from the NVIDIA Omniverse, an open computing platform for building and operating applications that connect with leading scientific computing visualization software. Working with the CAD files for the GA/DIII-D tokamak including equilibrium evolution as an exemplar tokamak application using Omniverse, the Princeton-NVIDIA collaboration has integrated modern AI/HPC-enabled near-real-time kinetic dynamics to connect and accelerate state-of-the-art, synthetic, HPC simulators to model fusion devices and control systems. The overarching goal is to deliver an interactive scientific digital twin of an advanced MFE tokamak that enables near-real-time simulation workflows built with Omniverse to eventually help open doors to new capabilities for generating clean power for a better future.