Gaussian boson sampling (GBS) is one of the leading approaches for demonstrating quantum computational advantage, but its application to practical real-world problems remains a central challenge. Here, we propose a GBS-based image recognition scheme inspired by extreme learning machine to enhance the performance of perceptron and implement it using our latest GBS device, jiǔzhāng 4.0. By avoiding in situ programmability of the photonic circuit, the scheme substantially reduces experimental overhead while retaining a large, fixed random feature map. Our approach utilizes an 8176-mode temporal-spatial hybrid encoding photonic processor, achieving approximately 2200 average photon clicks in the quantum computational advantage regime. We apply this scheme to classify images from the MNIST and Fashion-MNIST datasets, achieving a testing accuracy of 95.86% on MNIST and 85.95% on Fashion-MNIST, respectively. These results surpass both classical linear-kernel support vector machines and the three previous physical extreme-learning-machine experiments. In addition, we systematically explore the influence of three key hyperparameters and the efficiency of GBS in our experiments. Our results not only demonstrate the potential of GBS in real-world machine learning applications but will also inspire further advancements in powerful machine learning schemes utilizing GBS technology.

The toric code, when deformed in a way that preserves the self-duality ${\mathbb{Z}}_2$ symmetry exchanging the electric and magnetic excitations, admits a transition to a topologically trivial state that spontaneously breaks the ${\mathbb{Z}}_2$ symmetry. Numerically, this transition was found to be continuous, which makes it particularly enigmatic given the longstanding absence of a continuum field-theoretic description. In this Letter we propose such a continuum field theory for the transition dubbed the $SO(4{)}_{2,-2}$ Chern-Simons-Higgs (CSH) theory. We show that our field theory provides a natural “mean-field” understanding of the phase diagram. Moreover, it can be generalized to an entire series of theories, namely the $SO(4{)}_{k,-k}$ CSH theories, labeled by an integer $k$. For each $k>2$, the theory describes an analogous transition involving different non-Abelian topological orders, such as the double Fibonacci order ($k=3$) and the $S_3$ quantum double ($k=4$). For $k=1$, we conjecture that the corresponding CSH transition is in fact infrared dual to the $3d$ Ising transition, in close analogy with the particle-vortex duality of a complex scalar.

Bright entangled photon sources with on-demand quantum state control are essential for advancing quantum science and technologies. However, simultaneously achieving high brightness and state tunability remains challenging, as this necessitates a dual-phase control mechanism that facilitates both longitudinal phase matching and transverse polarization modulation. Here, we report a twin-twisted two-dimensional (2D) van der Waals (vdW) crystal architecture that enables spontaneous parametric down-conversion (SPDC) with both high generation efficiency and continuously tunable quantum states. The designed crystal comprises two sections of twisted 2D GaSe vdW flakes with opposite chiral handedness, in which interflake twist angle induces phase matching in SPDC, and intersection twisting provides an additional degree of freedom on tunability. We demonstrate that the twin-twisted GaSe crystals achieve a record-high photon-pair generation rate exceeding 13 200 Hz and a Bell state fidelity above 99% within 2D material family, and enable unprecedented tunability of quantum states from the maximally polarization entangled states to the fully separable states. This study opens a practical avenue toward twist-engineered quantum states, and paves the way for developing compact quantum photonic systems with tailored properties.

Matter-induced neutrino flavor mixing (the Mikheyev-Smirnov-Wolfenstein, or MSW, effect) is a central prediction of the neutrino mixing framework, but it has not been conclusively observed. Direct observation of the energy-dependent MSW transition in the solar electron-neutrino survival probability would solve this, but backgrounds have been prohibitive. We show that our new technique for suppressing muon-induced spallation backgrounds will allow JUNO to measure the MSW transition at $>4\sigma$ significance in 10 years. This would strongly support upcoming multibillion-dollar next-generation experiments and their goals in cementing the neutrino mixing framework.

We study the breakup dynamics of a two-dimensional strip of active chiral fluid using continuum hydrodynamics. Using slender body theory, we show that the breakup is driven by the interplay between chiral and surface tension forces, and that the minimum strip thickness decays to zero as a power law in finite time. The power law exponent and corresponding scaling function describing the shape of the strip near the break point are computed analytically, and shown to be in excellent agreement with numerical simulations of the underlying hydrodynamics. Qualitative agreement between experiment and simulation is also found.

We investigate $(k_1,k_2)$ extendibility of fermionic Gaussian states, a property central to quantum correlations and approximations of separability. We show that these states are $(k_1,k_2)$-extendible if and only if they admit a fermionic Gaussian extension, yielding a complete covariance-matrix characterization and a simple semidefinite program (SDP) whose size scales linearly with the number of modes. This provides necessary conditions for arbitrary fermionic states and is sufficient within the Gaussian setting. Our main result is a finite de Finetti–type theorem: we derive trace-norm bounds between $(k_1,k_2)$-extendible fermionic Gaussian states and separable states, improving previous exponential scaling to linear in the number of modes, with complementary relative entropy and squashed entanglement bounds. For two modes, upper and lower bounds match at order $1/\sqrt{k_1{k}_2}$. Extendibility also provides operational support for one of the different notions of separability in fermionic systems. Finally, for fermionic Gaussian channels, we provide a simple semidefinite program criterion for antidegradability and show that entanglement-breaking channels coincide with replacement channels, implying no nontrivial entanglement-breaking fermionic Gaussian channels exist.

We elucidate the boundary between classical and quantum computation by constructing qubit Clifford circuits with nonstabilizer inputs that can be efficiently simulated classically. We do so by casting the quantum circuits realizable by defect braiding in the surface code in terms of a Kirkwood-Dirac (KD) quasiprobability distribution, a generalization of a joint probability distribution. If this distribution remains a proper (positive) probability distribution throughout a circuit, then a classical algorithm can simulate the circuit efficiently. By leveraging recent results on the geometry of KD-positive states, we construct bound-magic states. Classical computers can efficiently simulate these bound-magic states’ evolutions under the circuits, although other magic states enable universal quantum computation when inputted. Furthermore, we show that KD nonpositivity is a resource monotone in this model. Thus, we establish KD nonpositivity as a necessary resource for quantum-computational advantages.

The flavor composition of cosmic neutrinos probes both the physical properties of their sources and oscillations over energies and distances unreachable by terrestrial experiments. Using 11.4 yr of IceCube data, we analyze the flavor composition of the all-sky neutrino flux from 5 TeV–10 PeV, significantly lowering the analysis threshold from a previous minimum of 35 TeV. The best-fit flavor fractions are $f_e,f_{\mu },f_{\tau }=0.30_{-0.15}^{+0.20}$, ${0.37}_{-0.16}^{+0.14}$, ${0.33}_{-0.25}^{+0.24}$. For the first time, we report that neutrinos of all three flavors are present in the observed cosmic neutrino flux at $>90\%$ confidence level. We infer the flavor composition at their sources, finding good agreement with pion decay while pure neutron decay is disfavored, with posterior probability $2.7\times {10}^{-5}$ ($\sim 4\sigma$).

We present observational constraints on large-scale white noise (LSWN) in the cosmic density field, a phenomenon predicted to arise from nonlinear mode coupling during cosmological evolution. Building on the theoretical framework of our companion paper, where we demonstrated that nonlinearities inevitably redistribute power from small to large scales through mode mixing, we confront these predictions with current cosmological data. We modify the class Boltzmann code to incorporate a white noise component $k_{\mathrm{BH}}/k$ in the primordial power spectrum and perform parameter estimation using current cosmological data. The nondetection of excess power on the largest observable scales places stringent upper bounds: $k_{\mathrm{BH}}\le 1.80\times {10}^{-13}{\mathrm{Mpc}}^{-1}$ at 99% confidence. These constraints imply the primordial power spectrum must deviate from a power law on small scales, perhaps with a sharp cutoff at $k_{\mathrm{cut}}\lesssim 0.03{\mathrm{pc}}^{-1}$ or through running of the spectral index with ${\alpha }_s\lesssim -0.019$. Our results demonstrate that LSWN provides a powerful probe of the primordial spectrum at scales orders of magnitude smaller than those directly observable, offering unique constraints on early-universe physics.

The formation of gravitationally bound overdensities of dark matter (DM), or halos, is a generic prediction of theories with particle DM. We present a measurement of halo properties in 28 quadruple image strong lens systems recently observed by JWST, and use these observations to constrain the free-streaming length, ${\lambda }_{\mathrm{FS}}$, of DM, a quantity that depends on the DM particle mass and formation mechanism. We improve on previous lensing analyses by simultaneously reconstructing extended lensed arcs with image positions and relative magnifications, enhancing sensitivity to perturbations by halos. Our analysis rules out deviations from the predictions of cold dark matter (CDM) on scales above ${10}^{7.2}{M}_{\odot }$ and ${10}^{7.4}{M}_{\odot }$ for subhalo abundance predicted by cosmological $N$-body simulations and semianalytic models, respectively. These bounds correspond to upper limits ${\lambda }_{\mathrm{FS}}<6.0\mathrm{kpc}$ and ${\lambda }_{\mathrm{FS}}<7.0\mathrm{kpc}$, and lower limits on the mass of a spin–1/2 thermal relic DM particle $m_{\text{therm}}>7.4\mathrm{keV}$ and $m_{\text{therm}}>6.5\mathrm{keV}$. Conversely, assuming a negligible free-streaming length, as predicted by CDM, we measure a projected mass in subhalos around elliptical galaxies ${1.7}_{-1.2}^{+2.6}\times {10}^7{\mathrm{M}}_{\odot }{\mathrm{kpc}}^{-2}$ at 95% confidence. These inferences confirm key predictions of the CDM paradigm.

The frequency-domain waveform emitted by a two-body scattering process is computed in the extreme-mass-ratio limit through the fifth post-Minkowskian (PM) order [i.e., $O(G^5)$] and the fractional sixth post-Newtonian (PN) order. The current accuracy of the scattering waveform obtained by quantum amplitude methods is the one-loop level corresponding to the 3PM order, whereas the 4PM waveform is known up to the 2PN order only as derived within the traditional multipolar-post-Minkowskian formalism. Direct comparison between these waveforms to the first order in the mass ratio shows that they differ at most by the effect of an angular-independent time shift, leading to a complete physical agreement at the same level of accuracy. The new results at the 4PM and 5PM orders thus provide a benchmark for future multiloop calculations. The waveform is also used to obtain the radiated energy at $O(G^6)$, extending it from 3PN to 6PN level.

We develop a consistent partially off-shell framework for evaluating higher-derivative actions of five-dimensional $\mathcal{N}=1$ gauged supergravity with Abelian vector multiplets on ${\mathrm{AdS}}_5$. Using the superconformal formalism, we show that the resulting holographic expression reproduces the trial $a$-anomaly coefficient of the dual conformal field theory, identifying the supergravity equations of motion with $a$ maximization. We present an exact correspondence between Chern-Simons couplings and the anomaly polynomial of the boundary theory. We illustrate our proposal by applying it to all known two- and four-derivative actions, including the “off-diagonal” invariants never before considered in the gauged supergravity literature. Finally, we argue that all invariants beyond four derivative yield no genuinely new contributions to asymptotically ${\mathrm{AdS}}_5$ BPS backgrounds, but instead reduce to linear combinations of the established two- and four-derivative actions.

Isotopic measurements of presolar silicon carbide grains from dying stars have revealed a puzzling overabundance of ${}^{94}\mathrm{Mo}$ that stellar nucleosynthesis models have failed to reproduce for two decades. This discrepancy challenged our understanding of the slow neutron-capture process ($s$-process) that forges approximately half of the elements heavier than iron. The key uncertainty lies at ${}^{94}\mathrm{Nb}$, a radioactive branching point where competition between neutron capture and beta decay governs the ${}^{94}\mathrm{Mo}$ production, yet the neutron-capture cross section had never been measured. Here, we report the first experimental determination of the ${}^{94}\mathrm{Nb}(\mathrm{n},\gamma ){}^{95}\mathrm{Nb}$ cross section important for Mo isotopic abundances. The measurement was enabled by a coordinated effort involving high-purity target preparation at Institute of Solid State and Materials Research Dresden, radioactive sample production at the Institut Laue-Langevin Grenoble, radiochemical characterization at Paul Scherrer Institut Villigen, and the time-of-flight CERN n_TOF facility using for the first time segmented total-energy detectors. Incorporation of the resulting Maxwellian-averaged cross section into fully coupled nucleosynthesis models of low-mass asymptotic giant-branch stars brings them into agreement with the presolar grain data. These results remove a major nuclear-physics input uncertainty at the ${}^{94}\mathrm{Nb}$ branching point and provide a firmer foundation for understanding the origin of ${}^{94}\mathrm{Mo}$ in the Solar System.