We present results from a systematic search for transiting short-period Giant Exoplanets around M-dwarf Stars (GEMS; P < 10 days, $R_p \gtrsim 8~R_\oplus$) within a distance-limited 100\,pc sample of 149,316 M-dwarfs using TESS-Gaia Light Curve (TGLC) data. This search led to the discovery of one new candidate GEM, following spectroscopic vetting of 12 additional candidates to eliminate astrophysical false positives and refine our occurrence rate estimates. We describe the development and application of the \texttt{TESS-miner} package and associated vetting procedures used in this analysis. To assess detection completeness, we conducted $\sim$ 72 million injection-recovery tests across $\sim$ 26,000 stars with an average of $\sim$3 sectors of data per star, subdivided into early-type (M0--M2.5), mid-type (M2.5--M4), and late-type (M4 or later) M-dwarfs. Our pipeline demonstrates high sensitivity across all M-dwarf subtypes within the injection bounds. We estimate the occurrence rates of short-period GEMS as a function of stellar mass, and combine our measured occurrence rates with those derived for FGK stars and fit an exponential trend with stellar mass, consistent with core-accretion theory predictions. We find GEMS occurrence rates of $0.067\% \pm 0.047\%$ for early-type M-dwarfs, $0.139\% \pm 0.069\%$ for mid-type, and $0.032\% \pm 0.032\%$ for late-type M-dwarfs, with a mean rate of $0.065^{+0.025}_{-0.027}\%$ across the full M-dwarf sample. We note that while our search spanned 1.0~\mathrm{days} < P < 10.0 days, these occurrence rates were calculated using planets orbiting with 1.0~\mathrm{days} < P < 5.0 days. This work lays the foundation for future occurrence rate investigations for GEMS.

While effective backdoor detection and inversion schemes have been developed for AIs used e.g. for images, there are challenges in "porting" these methods to LLMs. First, the LLM input space is discrete, which precludes gradient-based search over this space, central to many backdoor inversion methods. Second, there are ~30,000^k k-tuples to consider, k the token-length of a putative trigger. Third, for LLMs there is the need to blacklist tokens that have strong marginal associations with the putative target response (class) of an attack, as such tokens give false detection signals. However, good blacklists may not exist for some domains. We propose a LLM trigger inversion approach with three key components: i) discrete search, with putative triggers greedily accreted, starting from a select list of singletons; ii) implicit blacklisting, achieved by evaluating the average cosine similarity, in activation space, between a candidate trigger and a small clean set of samples from the putative target class; iii) detection when a candidate trigger elicits high misclassifications, and with unusually high decision confidence. Unlike many recent works, we demonstrate that our approach reliably detects and successfully inverts ground-truth backdoor trigger phrases.

Since black holes can be formed through widely varying processes, the horizon structure is highly complicated in the dynamical phase. Nonetheless, as numerical simulations show, the final state appears to be universal, well described by the Kerr geometry. How are all these large and widely varying deviations from the Kerr horizon washed out? To investigate this issue, we introduce a well-suited notion of horizon multipole moments and equations governing their dynamics, thereby providing a coordinate and slicing independent framework to investigate the approach to equilibrium. In particular, our flux formulas for multipoles can be used as analytical checks on numerical simulations and, in turn, the simulations could be used to fathom possible universalities in the way black holes approach their final equilibrium.

Linear Temporal Logic (LTL) offers a precise means for constraining the behavior of reinforcement learning agents. However, in many settings where both satisfaction and optimality conditions are present, LTL is insufficient to capture both. Instead, LTL-constrained policy optimization, where the goal is to optimize a scalar reward under LTL constraints, is needed. This constrained optimization problem proves difficult in deep Reinforcement Learning (DRL) settings, where learned policies often ignore the LTL constraint due to the sparse nature of LTL satisfaction. To alleviate the sparsity issue, we introduce Cycle Experience Replay (CyclER), a novel reward shaping technique that exploits the underlying structure of the LTL constraint to guide a policy towards satisfaction by encouraging partial behaviors compliant with the constraint. We provide a theoretical guarantee that optimizing CyclER will achieve policies that satisfy the LTL constraint with near-optimal probability. We evaluate CyclER in three continuous control domains. Our experimental results show that optimizing CyclER in tandem with the existing scalar reward outperforms existing reward-shaping methods at finding performant LTL-satisfying policies.

Transformer-based models, such as BERT and GPT, have been widely adopted in natural language processing (NLP) due to their exceptional performance. However, recent studies show their vulnerability to textual adversarial attacks where the model's output can be misled by intentionally manipulating the text inputs. Despite various methods that have been proposed to enhance the model's robustness and mitigate this vulnerability, many require heavy consumption resources (e.g., adversarial training) or only provide limited protection (e.g., defensive dropout). In this paper, we propose a novel method called dynamic attention, tailored for the transformer architecture, to enhance the inherent robustness of the model itself against various adversarial attacks. Our method requires no downstream task knowledge and does not incur additional costs. The proposed dynamic attention consists of two modules: (I) attention rectification, which masks or weakens the attention value of the chosen tokens, and (ii) dynamic modeling, which dynamically builds the set of candidate tokens. Extensive experiments demonstrate that dynamic attention significantly mitigates the impact of adversarial attacks, improving up to 33\% better performance than previous methods against widely-used adversarial attacks. The model-level design of dynamic attention enables it to be easily combined with other defense methods (e.g., adversarial training) to further enhance the model's robustness. Furthermore, we demonstrate that dynamic attention preserves the state-of-the-art robustness space of the original model compared to other dynamic modeling methods.

We develop a convolutional regularized least squares ($\texttt{CRLS}$) framework for reduced-order modeling of transonic flows with shocks. Conventional proper orthogonal decomposition (POD) based reduced models are attractive because of their optimality and low online cost; however, but they perform poorly when snapshots contain parameter-dependent discontinuities, leading to smeared shocks, stair-stepping, or non-physical oscillations. In $\texttt{CRLS}$, we first map each full-order snapshot to a smoother representation by applying a one-dimensional Gaussian convolution with reflect padding along the flow field coordinates. The convolution hyperparameters (kernel width and support) are selected automatically by Bayesian optimization on a held-out set of snapshots. POD bases are then extracted from the smoothed data, and the parametric dependence of the POD coefficients is learned via radial basis function interpolation. To recover sharp shock structures, we introduce an efficient deconvolution step formulated as a regularized least squares problem, where the regularization centers the reconstruction around a nearest-neighbor reference snapshot in parameter space. The resulting $\texttt{CRLS}$ surrogate is evaluated on inviscid transonic flow over the RAE2822 airfoil, modeled by the steady compressible Euler equations solved with SU2 over a Latin hypercube sample of Mach number and angle of attack. Compared with standard POD and smoothed-POD baselines, $\texttt{CRLS}$ yields markedly improved shock location and strength, lower surface-pressure and field-level errors, and a $42$\% reduction in the number of POD modes required to capture a fixed fraction of snapshot energy. These results demonstrate that $\texttt{CRLS}$ provides an accurate, data-efficient, and largely automated route to shock-aware reduced order models for high-speed aerodynamic design.

With the rapid development of Large Language Models (LLMs), we have witnessed intense competition among the major LLM products like ChatGPT, LLaMa, and Gemini. However, various issues (e.g. privacy leakage and copyright violation) of the training corpus still remain underexplored. For example, the Times sued OpenAI and Microsoft for infringing on its copyrights by using millions of its articles for training. From the perspective of LLM practitioners, handling such unintended privacy violations can be challenging. Previous work addressed the ``unlearning" problem of LLMs using gradient information, while they mostly introduced significant overheads like data preprocessing or lacked robustness. In this paper, contrasting with the methods based on first-order information, we revisit the unlearning problem via the perspective of second-order information (Hessian). Our unlearning algorithms, which are inspired by classic Newton update, are not only data-agnostic/model-agnostic but also proven to be robust in terms of utility preservation or privacy guarantee. Through a comprehensive evaluation with four NLP datasets as well as a case study on real-world datasets, our methods consistently show superiority over the first-order methods.

In the digital economy, technological innovations make it cheaper to produce high-quality content. For example, generative AI tools reduce costs for creators who develop content to be distributed online, but can also reduce production costs for the users who consume that content. These innovations can thus lead to disintermediation, since consumers may choose to use these technologies directly, bypassing intermediaries. To investigate when technological improvements lead to disintermediation, we study a game with an intermediary, suppliers of a production technology, and consumers. First, we show disintermediation occurs whenever production costs are too high or too low. We then investigate the consequences of disintermediation for welfare and content quality at equilibrium. While the intermediary is welfare-improving, the intermediary extracts all gains to social welfare and its presence can raise or lower content quality. We further analyze how disintermediation is affected by the level of competition between suppliers and the intermediary's fee structure. More broadly, our results take a step towards assessing how production technology innovations affect the survival of intermediaries and impact the digital economy.

Can stated preferences inform counterfactual analyses of actual choice? This research proposes a novel approach to researchers who have access to both stated choices in hypothetical scenarios and actual choices, matched or unmatched. The key idea is to use stated choices to identify the distribution of individual unobserved heterogeneity. If this unobserved heterogeneity is the source of endogeneity, the researcher can correct for its influence in a demand function estimation using actual choices and recover causal effects. Bounds on causal effects are derived in the case, where stated choice and actual choices are observed in unmatched data sets. These data combination bounds are of independent interest. We derive bootstrap inference for the bounds and show its good performance in a simulation experiment.

Graph Convolutional Networks (GCNs) are known to suffer from performance degradation as the number of layers increases, which is usually attributed to over-smoothing. Despite the apparent consensus, we observe that there exists a discrepancy between the theoretical understanding of over-smoothing and the practical capabilities of GCNs. Specifically, we argue that over-smoothing does not necessarily happen in practice, a deeper model is provably expressive, can converge to global optimum with linear convergence rate, and achieve very high training accuracy as long as properly trained. Despite being capable of achieving high training accuracy, empirical results show that the deeper models generalize poorly on the testing stage and existing theoretical understanding of such behavior remains elusive. To achieve better understanding, we carefully analyze the generalization capability of GCNs, and show that the training strategies to achieve high training accuracy significantly deteriorate the generalization capability of GCNs. Motivated by these findings, we propose a decoupled structure for GCNs that detaches weight matrices from feature propagation to preserve the expressive power and ensure good generalization performance. We conduct empirical evaluations on various synthetic and real-world datasets to validate the correctness of our theory.

We first show that the intrinsic, geometrical structure of a dynamical horizon is unique. A number of physically interesting constraints are then established on the location of trapped and marginally trapped surfaces in the vicinity of any dynamical horizon. These restrictions are used to prove several uniqueness theorems for dynamical horizons. Ramifications of some of these results to numerical simulations of black hole spacetimes are discussed. Finally several expectations on the interplay between isometries and dynamical horizons are shown to be borne out.

Final results are reported from operation of the PICO-60 C$_3$F$_8$ dark matter detector, a bubble chamber filled with 52 kg of C$_3$F$_8$ located in the SNOLAB underground laboratory. The chamber was operated at thermodynamic thresholds as low as 1.2 keV without loss of stability. A new blind 1404-kg-day exposure at 2.45 keV threshold was acquired with approximately the same expected total background rate as the previous 1167-kg-day exposure at 3.3 keV. This increased exposure is enabled in part by a new optical tracking analysis to better identify events near detector walls, permitting a larger fiducial volume. These results set the most stringent direct-detection constraint to date on the WIMP-proton spin-dependent cross section at 2.5 $\times$ 10$^{-41}$ cm$^2$ for a 25 GeV WIMP, and improve on previous PICO results for 3-5 GeV WIMPs by an order of magnitude.

Boundary conditions defining a generic isolated horizon are introduced. They generalize the notion available in the existing literature by allowing the horizon to have distortion and angular momentum. Space-times containing a black hole, itself in equilibrium but possibly surrounded by radiation, satisfy these conditions. In spite of this generality, the conditions have rich consequences. They lead to a framework, somewhat analogous to null infinity, for extracting physical information, but now in the \textit{strong} field regions. The framework also generalizes the zeroth and first laws of black hole mechanics to more realistic situations and sheds new light on the `origin' of the first law. Finally, it provides a point of departure for black hole entropy calculations in non-perturbative quantum gravity.

Geometrical structures intrinsic to non-expanding, weakly isolated and isolated horizons are analyzed and compared with structures which arise in other contexts within general relativity, e.g., at null infinity. In particular, we address in detail the issue of singling out the preferred normals to these horizons required in various applications. This work provides powerful tools to extract invariant, physical information from numerical simulations of the near horizon, strong field geometry. While it complements the previous analysis of laws governing the mechanics of weakly isolated horizons, prior knowledge of those results is not assumed.

The development of NekRS, a GPU-oriented thermal-fluids simulation code based on the spectral element method (SEM) is described. For performance portability, the code is based on the open concurrent compute abstraction and leverages scalable developments in the SEM code Nek5000 and in libParanumal, which is a library of high-performance kernels for high-order discretizations and PDE-based miniapps. Critical performance sections of the Navier-Stokes time advancement are addressed. Performance results on several platforms are presented, including scaling to 27,648 V100s on OLCF Summit, for calculations of up to 60B gridpoints.

Loop quantum gravity is an approach to quantum gravity that starts from the Hamiltonian formulation in terms of a connection and its canonical conjugate. Quantization proceeds in the spirit of Dirac: First one defines an algebra of basic kinematical observables and represents it through operators on a suitable Hilbert space. In a second step, one implements the constraints. The main result of the paper concerns the representation theory of the kinematical algebra: We show that there is only one cyclic representation invariant under spatial diffeomorphisms. While this result is particularly important for loop quantum gravity, we are rather general: The precise definition of the abstract *-algebra of the basic kinematical observables we give could be used for any theory in which the configuration variable is a connection with a compact structure group. The variables are constructed from the holonomy map and from the fluxes of the momentum conjugate to the connection. The uniqueness result is relevant for any such theory invariant under spatial diffeomorphisms or being a part of a diffeomorphism invariant theory.

The NEID Earth Twin Survey (NETS) has been delivering a rich set of precise radial velocity (RV) measurements for 41 bright, nearby main sequence stars. Here, we describe the status of the survey after three years on sky and we present the full set of RV measurements and accompanying stellar activity indicators. We discuss intermediate survey diagnostics, including calibration of the known RV zero point offset introduced following the Contreras fire in 2022 and the identification of an undiagnosed and previously unknown zero point offset in 2021. An analysis of our data set using RVSearch demonstrates that for these target stars, NEID is independently sensitive to nearly all known planets with periods shorter than the NETS observing baseline. We also highlight a number of newly detected RV signals, which present exciting opportunities for future investigations.