Leptoquark models are prime candidates for new physics (NP) explanations of the long-standing anomalies in semi-leptonic $B$ decays; $b\to c \tau \bar\nu$ (encoded in $R(D^{(\ast)})$) and $b\to s\ell\bar\ell (\ell=e,\mu)$ transitions. Furthermore, Belle II and NA62 reported weaker-than-expected limits on $B^+ \to K^+ \nu\bar\nu$ and $K^+ \to \pi^+ \nu \bar\nu$, respectively. While the $R(D^{(\ast)})$ and $b\to s\ell \bar\ell$ measurements can be explained with NP contributions at the $O(10\%)$ level, the neutrino channels suggest that the NP effect could be comparable in size to the Standard Model one. In this context, we consider the two types of leptoquark models with minimal sets of the couplings that can best describe the semi-leptonic $B$ anomalies and lead at the same time to effects in the neutrino modes, the singlet-triplet scalar leptoquark model ($S_1+S_3$) and the singlet vector leptoquark model ($U_1$). More specifically, the neutrino channels pose non-trivial constraints on the parameter space, and we find that large effects (i.e., accounting for the current central value) in $B\to K^{(*)}\nu\bar\nu$ are only possible in the $S_1+S_3$ setup, while both models can account for the central value of $K^+\to \pi^+\nu\bar\nu$.

Not identical but similar objects are ubiquitous in our world, ranging from four-legged animals such as dogs and cats to cars of different models and flowers of various colors. This study addresses a novel task of matching such non-identical objects at the pixel level. We propose a weighting scheme of descriptors that incorporates semantic information from object detectors into existing sparse image matching methods, extending their targets from identical objects captured from different perspectives to semantically similar objects. The experiments show successful matching between non-identical objects in various cases, including in-class design variations, class discrepancy, and domain shifts (e.g., photo--drawing and image corruptions).

Offline reinforcement learning, which learns solely from datasets without environmental interaction, has gained attention. This approach, similar to traditional online deep reinforcement learning, is particularly promising for robot control applications. Nevertheless, its robustness against real-world challenges, such as joint actuator faults in robots, remains a critical concern. This study evaluates the robustness of existing offline reinforcement learning methods using legged robots from OpenAI Gym based on average episodic rewards. For robustness evaluation, we simulate failures by incorporating both random and adversarial perturbations, representing worst-case scenarios, into the joint torque signals. Our experiments show that existing offline reinforcement learning methods exhibit significant vulnerabilities to these action perturbations and are more vulnerable than online reinforcement learning methods, highlighting the need for more robust approaches in this field.

We explore the possibility that the interactions of a light scalar singlet, which mixes with the Standard Model~(SM) Higgs, also receive other UV contributions of comparable size. We focus, in particular, on the flavor aligned limit, where couplings of the light scalar to the SM are almost flavor diagonal, but not necessarily proportional to the Higgs Yukawa couplings. The phenomenology of such a general flavor aligned light scalar differs from both the Higgs-mixed scalar, as well as from a general axion-like particle. We explore this for light scalar masses below a few hundred MeV, such that they can be produced in kaon decays, and in decays of $\eta$ and $\eta'$ mesons, and the transitions described using chiral perturbation theory. We then derive constraints on the light scalar interactions, assuming that light scalar decays are either just into photons or are invisible. We also discuss several UV examples of such light scalar models: a two-Higgs doublet model extended by a light scalar, a light dilaton from the dark sector, and a SM extended by heavy vector-like quarks and a light scalar. For the latter we also performed matching onto low energy theory at one-loop.

We point out that using current knowledge of ${\cal B}(K^0_L\to\mu^+\mu^-)$ and ${\cal B}(K^0_L\to \gamma\gamma)$, one can extract short-distance information from the combined measurement of the time-integrated CP asymmetry, $A_{\rm CP}(K^0\to\mu^+\mu^-)$, and of ${\cal B}(K^0_S\to\mu^+\mu^-)$. We discuss the interplay between this set of observables, and demonstrate that determining ${\rm sign}[A_{\rm CP}(K^0\to\mu^+\mu^-)]$ would eliminate the discrete ambiguity in the Standard Model prediction for ${\cal B}(K^0_L\to\mu^+\mu^-)$. We then move on to feasibility studies within an LHCb-like setup, using both time-integrated and time-dependent information, employing $K^0$ and $\overline K{}^0$ tagging methods. We find that, within an optimistic scenario, the short-distance amplitude, proportional to the CKM parameter combination $|A^2\lambda^5\bar\eta|$, could be constrained by LHCb at the level of about $35\%$ of its Standard Model value, and the discrete ambiguity in ${\cal B}(K^0_L\to\mu^+\mu^-)_{\rm SM}$ could be resolved at more than $3\sigma$ by the end of the high luminosity LHC.

We study the renormalization of the QCD $\theta$ angle at the two-loop level focusing on divergent and finite $CP$-violating contributions from evanescent operators, using dimensional regularization with the BMHV scheme. When one considers the Lagrangian in $d$-dimensional space-time instead of four dimensions in dimensional regularization, evanescent operators that break the chiral symmetry are induced. Consequently, $T$-odd and $P$-odd fermion loops in the BMHV scheme generate evanescent contributions to the QCD $\theta$ angle. We carefully classify the evanescent contributions into two types: one originating from the evanescent operators at the one-loop level and the other directly produced by two-loop calculations. We show that renormalization of the parameters in the BMHV scheme keeps removing those unphysical contributions to the QCD $\theta$ angle at any scale at the two-loop level. We also discuss how the rephasing invariance of the radiative correction to the QCD $\theta$ angle is realized in the BMHV scheme.

Recent progresses in the measurements of the electric dipole moment (EDM) of the electron using the paramagnetic atom or molecule are remarkable. In this paper, we calculate a contribution to the electron EDM at three-loop level, introducing the CP-violating Yukawa couplings of new SU(2)$_L$ multiplets. At two-loop level, the Yukawa interactions generate a CP-violating dimension-six operator, composed of three SU(2)$_L$ field strengths, called the electroweak Weinberg operator. Another one-loop diagram with the operator inserted induces the electron EDM. We find that even if new SU(2)$_L$ particles have masses around TeV-scale, the electron EDM may be larger than the Standard Model contribution to the paramagnetic atom or molecule EDMs. We also discuss the relation between the Barr-Zee diagram contribution at two-loop level and the three-loop one, assuming that the SM Higgs has new Yukawa interactions with the SU(2)$_L$ multiplets.

Quantum-like modeling (QLM) - quantum theory applications outside of physics - are intensively developed with applications in biology, cognition, psychology, and decision-making. For cognition, QLM should be distinguished from quantum reductionist models in the spirit of Hameroff and Penrose and well as Umezawa and Vitiello. QLM is not concerned with just quantum physical processes in the brain but also QL information processing by macroscopic neuronal structures. Although QLM of cognition and decision-making has seen some success, it suffers from a knowledge gap that exists between oscillatory neuronal network functioning in the brain and QL behavioral patterns. Recently, steps toward closing this gap have been taken using the generalized probability theory and prequantum classical statistical field theory (PCSFT) - a random field model beyond the complex Hilbert space formalism. PCSFT is used to move from the classical ``oscillatory cognition'' of the neuronal networks to QLM for this http URL. In this study, we addressed the most difficult problem within this construction: QLM for entanglement generation by classical networks, i.e., mental entanglement. We started with the observational approach to entanglement based on operator algebras describing local observables and bringing into being the tensor product structure in the space of QL states. Moreover, we applied the standard states entanglement approach: entanglement generation by spatially separated networks in the brain. Finally, we discussed possible future experiments on mental entanglement detection using the EEG/MEG technique.

Rest-frame optical observations with the James Webb Space Telescope (JWST) have uncovered a population of massive galaxies, exceeding $10^{10}$ solar masses, present less than a billion years after the Big Bang. The large stellar masses of these galaxies require an efficient conversion of baryons into stars, which may exceed theoretical expectations. However, the formation process of massive galaxies so early in the Universe's history is perplexing, as observations provide limited information to constrain their evolutionary pathways. Here, we present multi-wavelength observations of a galaxy complex consisting of at least five galaxies within a $\sim10\,{\rm kpc}$ region, referred to as the \quintet, using JWST and the Atacama Large Millimeter/submillimeter Array. This system, located in the core of a galaxy protocluster at approximately 650 million years after the Big Bang, reveals the detailed physical processes involved in the formation of massive galaxies. These processes include a dynamic cycles of merger induced gas stripping, leading the temporal termination of star formation, and recycling of the stripped gas, with subsequent enhancement of star formation in other galaxies of the system, which is expected to evolve into massive galaxies that host more than $10^{10}$ solar masses of stars. The new observations represent the first comprehensive evidence of a massive galaxy formation through gas-rich, multiple-galaxy mergers induced by a dense protocluster environment in the $650\,{\rm Myrs}$ after the Big Bang. Our results suggest that the protocluster core is indeed one of the main drivers of efficient galaxy formation and rapid evolution in the early Universe, as predicted by theoretical studies.

Highly magnified individual stars such as Icarus and Earendel have recently been observed near critical curves of galaxy clusters with Hubble Space Telescope (HST) observations. These stars are estimated to be magnified with a factor of more than a few thousands. In addition to the smooth mass distribution in the macro-lens model, the distribution of microlenses originating from, for instance, intracluster stars affects the event rate and the peak magnification significantly. We propose an analytic model of the high-magnification tail of the probability distribution function (PDF) in which the probability is assumed to be proportional to the number of independent microlens critical curves. Our model can explain the parameter dependence of the PDF on the mass fraction of the microlenses and the background magnification seen in ray-tracing simulations. The effect of a finite source size is also studied to derive a fitting formula for the suppression factor. For an application of our model, we calculate the event rate of the Icarus-like system and the probability distribution of observed positions of such system, showing good agreement with the HST observations. Our model predicts a complicated dependence of the probability distribution of observed positions of highly magnified events on the magnification threshold.

To better understand the behavior of image classifiers, it is useful to visualize the contribution of individual pixels to the model prediction. In this study, we propose a method, MoXI ($\textbf{Mo}$del e$\textbf{X}$planation by $\textbf{I}$nteractions), that efficiently and accurately identifies a group of pixels with high prediction confidence. The proposed method employs game-theoretic concepts, Shapley values and interactions, taking into account the effects of individual pixels and the cooperative influence of pixels on model confidence. Theoretical analysis and experiments demonstrate that our method better identifies the pixels that are highly contributing to the model outputs than widely-used visualization by Grad-CAM, Attention rollout, and Shapley value. While prior studies have suffered from the exponential computational cost in the computation of Shapley value and interactions, we show that this can be reduced to quadratic cost for our task. The code is available at https://github.com/KosukeSumiyasu/MoXI.

The IceCube Neutrino Observatory is a cubic-kilometer-scale high-energy neutrino detector built into the ice at the South Pole. Construction of IceCube, the largest neutrino detector built to date, was completed in 2011 and enabled the discovery of high-energy astrophysical neutrinos. We describe here the design, production, and calibration of the IceCube digital optical module (DOM), the cable systems, computing hardware, and our methodology for drilling and deployment. We also describe the online triggering and data filtering systems that select candidate neutrino and cosmic ray events for analysis. Due to a rigorous pre-deployment protocol, 98.4% of the DOMs in the deep ice are operating and collecting data. IceCube routinely achieves a detector uptime of 99% by emphasizing software stability and monitoring. Detector operations have been stable since construction was completed, and the detector is expected to operate at least until the end of the next decade.

We present cosmology results from a blinded joint analysis of cosmic shear, $\xi_{\pm}(\vartheta)$, galaxy-galaxy weak lensing, $\Delta\!\Sigma(R)$, and projected galaxy clustering, $w_{\rm p}(R)$, measured from the Hyper Suprime-Cam three-year (HSC-Y3) shape catalog and the Sloan Digital Sky Survey (SDSS) DR11 spectroscopic galaxy catalog - a 3$\times$2pt cosmology analysis. We define luminosity-cut samples of SDSS galaxies to serve as the tracers of $w_{\rm p}$ and as the lens samples for $\Delta\!\Sigma$ in three spectroscopic redshift bins spanning the range $0.15

Applying machine learning to real-world medical data, e.g. from hospital archives, has the potential to revolutionize disease detection in brain images. However, detecting pathology in such heterogeneous cohorts is a difficult challenge. Normative modeling, a form of unsupervised anomaly detection, offers a promising approach to studying such cohorts where the ``normal'' behavior is modeled and can be used at subject level to detect deviations relating to disease pathology. Diffusion models have emerged as powerful tools for anomaly detection due to their ability to capture complex data distributions and generate high-quality images. Their performance relies on image restoration; differences between the original and restored images highlight potential abnormalities. However, unlike normative models, these diffusion model approaches do not incorporate clinical information which provides important context to guide the disease detection process. Furthermore, standard approaches often poorly restore healthy regions, resulting in poor reconstructions and suboptimal detection performance. We present CADD, the first conditional diffusion model for normative modeling in 3D images. To guide the healthy restoration process, we propose a novel inference inpainting strategy which balances anomaly removal with retention of subject-specific features. Evaluated on three challenging datasets, including clinical scans, which may have lower contrast, thicker slices, and motion artifacts, CADD achieves state-of-the-art performance in detecting neurological abnormalities in heterogeneous cohorts.

Textual explanations make image classifier decisions transparent by describing the prediction rationale in natural language. Large vision-language models can generate captions but are designed for general visual understanding, not classifier-specific reasoning. Existing zero-shot explanation methods align global image features with language, producing descriptions of what is visible rather than what drives the prediction. We propose TEXTER, which overcomes this limitation by isolating decision-critical features before alignment. TEXTER identifies the neurons contributing to the prediction and emphasizes the features encoded in those neurons -- i.e., the decision-critical features. It then maps these emphasized features into the CLIP feature space to retrieve textual explanations that reflect the model's reasoning. A sparse autoencoder further improves interpretability, particularly for Transformer architectures. Extensive experiments show that TEXTER generates more faithful and interpretable explanations than existing methods. The code will be publicly released.