Search

Autonomous agents powered by large language models (LLMs) have the potential to enhance human capabilities, assisting with digital tasks from sending emails to performing data analysis. The abilities of existing LLMs at such tasks are often hindered by the lack of high-quality agent data from the corresponding environments they interact with. We propose Learn-by-interact, a data-centric framework to adapt LLM agents to any given environments without human annotations. Learn-by-interact synthesizes trajectories of agent-environment interactions based on documentations, and constructs instructions by summarizing or abstracting the interaction histories, a process called backward construction. We assess the quality of our synthetic data by using them in both training-based scenarios and training-free in-context learning (ICL), where we craft innovative retrieval approaches optimized for agents. Extensive experiments on SWE-bench, WebArena, OSWorld and Spider2-V spanning across realistic coding, web, and desktop environments show the effectiveness of Learn-by-interact in various downstream agentic tasks -- baseline results are improved by up to 12.2\% for ICL with Claude-3.5 and 19.5\% for training with Codestral-22B. We further demonstrate the critical role of backward construction, which provides up to 14.0\% improvement for training. Our ablation studies demonstrate the efficiency provided by our synthesized data in ICL and the superiority of our retrieval pipeline over alternative approaches like conventional retrieval-augmented generation (RAG). We expect that Learn-by-interact will serve as a foundation for agent data synthesis as LLMs are increasingly deployed at real-world environments.

Conversational assistants are increasingly popular across diverse real-world applications, highlighting the need for advanced multimodal speech modeling. Speech, as a natural mode of communication, encodes rich user-specific characteristics such as speaking rate and pitch, making it critical for effective interaction. Our work introduces a data-centric customization approach for efficiently enhancing multimodal understanding in conversational speech modeling. Central to our contributions is a novel multi-task learning paradigm that involves designing auxiliary tasks to utilize a small amount of speech data. Our approach achieves state-of-the-art performance on the Spoken-SQuAD benchmark, using only 10% of the training data with open-weight models, establishing a robust and efficient framework for audio-centric conversational modeling. We also introduce ASK-QA, the first dataset for multi-turn spoken dialogue with ambiguous user requests and dynamic evaluation inputs. Code and data forthcoming., Comment: 22 pages, 6 figures, 14 tables

We study a chain of interacting individual quantum systems connected to heat baths at different temperatures on both ends. Starting with the two-system case, we thoroughly investigate the conditions for heat rectification (asymmetric heat conduction), compute thermal conductance, and generalize the results to longer chains. We find that heat rectification in the weak coupling regime can be independent of the chain length and that negative differential thermal conductance occurs. We also examine the relationship between heat rectification with entanglement and the entropy production. In the strong coupling regime, the system exhibits an asymmetric Rabi-type splitting in the thermal conductance, leading to enhanced heat transport and improved rectification inaccessible in the weak coupling. This setup represents the simplest quantum thermal machine that consumes incoherent resources and delivers entanglement while acting as a rectifier and heat valve., Comment: 13 pages, 7 figures

Noisy voltage sources can be a limiting factor for fundamental physics experiments as well as for device applications in quantum information, mesoscopic circuits, magnetometry, and other fields. The best commercial DC voltage sources can be programmed to approximately six digits and have intrinsic noise in the microvolt range. On the other hand the noise level in metrological Josephson-junction based voltage standards is sub-femtovolt. Although such voltage standards can be considered "noiseless," they are generally not designed for continuous tuning of the output voltage nor for supplying current to a load at cryogenic temperatures. We propose a Josephson effect based voltage source, as opposed to a voltage standard, operating in the 30-160 uV range which can supply over 100 nA of current to loads at milli-Kelvin temperatures. We describe the operating principle, the sample design, and the calibration procedure to obtain continuous tunability. We show current-voltage characteristics of the device, demonstrate how the voltage can be adjusted without DC control connections to room-temperature electronics, and showcase an experiment coupling the source to a mesoscopic load, a small Josephson junction. Finally we characterize the performance of our source by measuring the voltage noise at the load, 50 pV RMS, which is attributed to parasitic resistances in the cabling. This work establishes the use of the Josephson effect for voltage biasing extremely sensitive quantum devices., Comment: 9 pages, 8 figures

Semiconductor components based on silicon carbide (SiC) are a key component for high-power electronics. Their behavior is determined by the interplay of charges and electric fields, which is typically described by modeling and simulations that are calibrated by nonlocal electric properties. So far, there are no experimental methods that allow for the 3D mapping of both the electric field and the concentrations of free charge carriers inside an electronic device. To fulfill this information gap, we propose an operando method that utilizes single silicon vacancy (VSi) centers in 4H-SiC. The VSi centers are at various positions in the intrinsic region of a pin-diode. To monitor the local static electric field, we perform Stark shift measurements based on photoluminescence excitation (PLE), which allows us to infer the expansion of the depletion zone and therefore to determine the local concentration of dopants. Besides this, we show that our measurements allow us to additionally obtain the local concentration of free charge carriers. The method presented here therefore paves the way for a new quantum-enhanced electronic device technology, capable of mapping the interplay of mobile charges and electric fields in a working semiconductor device with nanometer precision.

Retrieval-Augmented Generation (RAG), while effective in integrating external knowledge to address the limitations of large language models (LLMs), can be undermined by imperfect retrieval, which may introduce irrelevant, misleading, or even malicious information. Despite its importance, previous studies have rarely explored the behavior of RAG through joint analysis on how errors from imperfect retrieval attribute and propagate, and how potential conflicts arise between the LLMs' internal knowledge and external sources. We find that imperfect retrieval augmentation might be inevitable and quite harmful, through controlled analysis under realistic conditions. We identify the knowledge conflicts between LLM-internal and external knowledge from retrieval as a bottleneck to overcome in the post-retrieval stage of RAG. To render LLMs resilient to imperfect retrieval, we propose Astute RAG, a novel RAG approach that adaptively elicits essential information from LLMs' internal knowledge, iteratively consolidates internal and external knowledge with source-awareness, and finalizes the answer according to information reliability. Our experiments using Gemini and Claude demonstrate that Astute RAG significantly outperforms previous robustness-enhanced RAG methods. Notably, Astute RAG is the only approach that matches or exceeds the performance of LLMs without RAG under worst-case scenarios. Further analysis reveals that Astute RAG effectively resolves knowledge conflicts, improving the reliability and trustworthiness of RAG systems., Comment: Preprint

The electron antinuetrino flux limits are presented for the brightest gamma-ray burst (GRB) of all time, GRB221009A, over a range of 1.8-200 MeV using the Kamioka Liquid Scintillator Anti Neutrino Detector (KamLAND). Using a variety of time windows to search for electron antineutrinos coincident with the GRB, we set an upper limit on the flux under the assumption of various neutrino source spectra. No excess was observed in any time windows ranging from seconds to days around the event trigger time. The limits are compared to the results presented by IceCube., Comment: 8 pages, 3 figures, 1 table

In this study, we explore the entanglement and correlation in two-nucleon systems using isospin formalism. With the help of Slater decomposition, we derive analytical expressions for various entanglement measures. Specifically, we analyse the one- and two-mode entropies, mutual informations, and a basis-independent characteristic known as the one-body entanglement entropy. To understand the impact of pairing, we consider interactions involving isovector and isoscalar L=0 pairing terms. Our findings show that certain pairing interactions can maximize one-body entanglement entropy of ground states when both total angular momentum and total isospin have zero projections. We provide numerical examples for the sd shell and explore the mutual informations in LS coupled and jj coupled single-particle bases. We find that the shell structure and angular momentum coupling significantly impact the measures of entanglement. We outline the implications of conserving angular momentum and isospin on one-mode entropies, irrespective of particle number., Comment: 32 pages, 6 figures

In this paper, we develop a class of interacting particle Langevin algorithms to solve inverse problems for partial differential equations (PDEs). In particular, we leverage the statistical finite elements (statFEM) formulation to obtain a finite-dimensional latent variable statistical model where the parameter is that of the (discretised) forward map and the latent variable is the statFEM solution of the PDE which is assumed to be partially observed. We then adapt a recently proposed expectation-maximisation like scheme, interacting particle Langevin algorithm (IPLA), for this problem and obtain a joint estimation procedure for the parameters and the latent variables. We consider three main examples: (i) estimating the forcing for linear Poisson PDE, (ii) estimating the forcing for nonlinear Poisson PDE, and (iii) estimating diffusivity for linear Poisson PDE. We provide computational complexity estimates for forcing estimation in the linear case. We also provide comprehensive numerical experiments and preconditioning strategies that significantly improve the performance, showing that the proposed class of methods can be the choice for parameter inference in PDE models., Comment: 23 pages, 6 figures

Photons emitted from high-energy electron beam interactions with high-field systems, such as the upcoming FACET-II experiments at SLAC National Accelerator Laboratory, may provide deep insight into the electron beam's underlying dynamics at the interaction point. With high-energy photons being utilized to generate electron-positron pairs in a novel spectrometer, there remains a key problem of interpreting the spectrometer's raw data to determine the energy distribution of the incoming photons. This paper uses data from simulations of the primary radiation emitted from electron interactions with a high-field, short-pulse laser to determine optimally reliable methods of reconstructing the measured photon energy distributions. For these measurements, recovering the emitted 10 MeV to 10 GeV photon energy spectra from the pair spectrometer currently being commissioned requires testing multiple methods to finalize a pipeline from the spectrometer data to incident photon and, by extension, electron beam information. In this study, we compare the performance QR decomposition, a matrix deconstruction technique and neural network with and without maximum likelihood estimation (MLE). Although QR decomposition proved to be the most effective theoretically, combining machine learning and MLE proved to be superior in the presence of noise, indicating its promise for analysis pipelines involving high-energy photons., Comment: 11 Pages, 12 Figures. arXiv admin note: substantial text overlap with arXiv:2209.12119

Deep neural networks are vulnerable to adversarial examples, and adversarial attacks that generate adversarial examples have been studied in this context. Existing studies imply that increasing the diversity of model outputs contributes to improving the attack performance. This study focuses on the Auto Conjugate Gradient (ACG) attack, which is inspired by the conjugate gradient method and has a high diversification performance. We hypothesized that increasing the distance between two consecutive search points would enhance the output diversity. To test our hypothesis, we propose Rescaling-ACG (ReACG), which automatically modifies the two components that significantly affect the distance between two consecutive search points, including the search direction and step size. ReACG showed higher attack performance than that of ACG, and is particularly effective for ImageNet models with several classification classes. Experimental results show that the distance between two consecutive search points enhances the output diversity and may help develop new potent attacks. The code is available at \url{https://github.com/yamamura-k/ReACG}, Comment: ICPRAI2024

We explore the utility of quantum spin squeezing in quantum magnetometry, focusing on three-level (qutrit) Nitrogen-Vacancy (NV) centers within diamond, utilizing a standard Ramsey interferometry pulse protocol. Our investigation incorporates the effects of dephasing and relaxation on NV centers' dynamics during Ramsey measurements, modeled via the Lindblad quantum master equation. We conduct a comparative analysis between the metrological capabilities of a single NV center and a pair of NV centers, considering Quantum Fisher Information both with and without spin squeezing. The quantum correlations between NV centers are assessed through the evaluation of the Kitagawa-Ueda spin squeezing parameter within a two-level manifold. Additionally, parallel calculations are conducted using a two-level model (qubit) for NV centers. Our findings reveal that leveraging qutrits and spin squeezing yields enhanced magnetometric precision, albeit constrained by dephasing effects. Nevertheless, even in the absence of dynamical decoupling methods to mitigate environmental noise, strategic timing of squeezing and free evolution can sustain the advantages of qutrit-based magnetometry.

A novel method to reconstruct the energy of hadronic showers in the CMS High Granularity Calorimeter (HGCAL) is presented. The HGCAL is a sampling calorimeter with very fine transverse and longitudinal granularity. The active media are silicon sensors and scintillator tiles readout by SiPMs and the absorbers are a combination of lead and Cu/CuW in the electromagnetic section, and steel in the hadronic section. The shower reconstruction method is based on graph neural networks and it makes use of a dynamic reduction network architecture. It is shown that the algorithm is able to capture and mitigate the main effects that normally hinder the reconstruction of hadronic showers using classical reconstruction methods, by compensating for fluctuations in the multiplicity, energy, and spatial distributions of the shower's constituents. The performance of the algorithm is evaluated using test beam data collected in 2018 prototype of the CMS HGCAL accompanied by a section of the CALICE AHCAL prototype. The capability of the method to mitigate the impact of energy leakage from the calorimeter is also demonstrated.

We apply a recently proposed computational protocol for a neural-network-supported configuration interaction (NN CI) calculation to the paradigmatic N$_2$ molecule. By comparison of correlation energy, binding energy, and the full dissociation curve to experimental and full CI benchmarks, we demonstrate the applicability and robustness of our approach for the first time in the context of molecular systems, and offer thereby a new complementary tool in the family of machine-learning-based computation methods. The main advantage of the method lies in the efficiency of the neural-network-selected many-body basis set. Specifically, we approximate full CI results obtained on bases of $\approx 10^{10}$ Slater Determinants with only $\approx10^{5}$ determinants with good accuracy. The high efficiency of the NN CI approach underlines its potential for broader applications such as structural optimizations and even computation of spectroscopic observables in systems for which computational resources are a limiting factor., Comment: 8 pages, 4 figures

Addressing the challenge of effectively processing long contexts has become a critical issue for Large Language Models (LLMs). Two common strategies have emerged: 1) reducing the input length, such as retrieving relevant chunks by Retrieval-Augmented Generation (RAG), and 2) expanding the context window limit of LLMs. However, both strategies have drawbacks: input reduction has no guarantee of covering the part with needed information, while window extension struggles with focusing on the pertinent information for solving the task. To mitigate these limitations, we propose Chain-of-Agents (CoA), a novel framework that harnesses multi-agent collaboration through natural language to enable information aggregation and context reasoning across various LLMs over long-context tasks. CoA consists of multiple worker agents who sequentially communicate to handle different segmented portions of the text, followed by a manager agent who synthesizes these contributions into a coherent final output. CoA processes the entire input by interleaving reading and reasoning, and it mitigates long context focus issues by assigning each agent a short context. We perform comprehensive evaluation of CoA on a wide range of long-context tasks in question answering, summarization, and code completion, demonstrating significant improvements by up to 10% over strong baselines of RAG, Full-Context, and multi-agent LLMs., Comment: 19 pages, 6 figures

Large language models (LLMs) aligned through reinforcement learning from human feedback (RLHF) have quickly become one of the dominant paradigms for building intelligent conversational assistant agents. However, despite their strong performance across many benchmarks, LLM-based agents still lack conversational skills such as disambiguation: when generalized assistants are faced with ambiguity, they often overhedge or implicitly guess users' ground-truth intents rather than asking clarification questions, and under task-specific settings, high-quality conversation samples are often limited, affecting models' ability to learn optimal dialogue action policies. We propose Action-Based Contrastive Self-Training (henceforth ACT), a quasi-online preference optimization algorithm based on Direct Preference Optimization (DPO) which allows for sample-efficient dialogue policy learning in multi-turn conversation. We demonstrate ACT's efficacy under sample-efficient conditions in three difficult conversational tasks: tabular-grounded question-answering, machine reading comprehension, and AmbigSQL, a novel task for disambiguating information-seeking requests for text-to-SQL generation. Additionally, we propose evaluating LLMs' ability to function as conversational agents by examining whether they can implicitly recognize and reason about ambiguity in conversation. ACT demonstrates substantial conversation modeling improvements over standard approaches to supervised fine-tuning and DPO.

Despite their significant advancements, Multimodal Large Language Models (MLLMs) often generate factually inaccurate information, referred to as hallucination. In this work, we address object hallucinations in MLLMs, where information is generated about an object not present in the input image. We introduce Data-augmented Phrase-level Alignment (DPA), a novel loss which can be applied to instruction-tuned off-the-shelf MLLMs to mitigate hallucinations, while preserving their general vision-language capabilities. To fine-tune MLLMs with DPA, we first generate a set of `hallucinated' and `correct' response pairs through generative data augmentation by selectively altering the ground-truth information of the correct responses at a phrase level. The DPA loss is then used to train MLLMs to reduce the likelihood of hallucinated phrases compared to the correct ones. Our thorough evaluation on various benchmarks confirms the effectiveness of DPA in mitigating hallucination while retaining the out-of-the-box performance of the MLLMs on general tasks. For instance, MLLMs finetuned with DPA, which we refer to as Hallucination Attenuated Language and Vision Assistant (HALVA), improve F1 by up to 13.4% on hallucination visual question-answering and reduce the hallucination rate by up to 4.2% on image description tasks., Comment: 45 pages

In this work, we explore the strain and curvature effects on the electronic properties of a curved graphene structure, called the graphene wormhole. The electron dynamics is described by a massless Dirac fermion containing position--dependent Fermi velocity. In addition, the strain produces a pseudo--magnetic vector potential to the geometric coupling. For an isotropic strain tensor, the decoupled components of the spinor field exhibit a supersymmetric (SUSY) potential, depending on the centrifugal term and the external magnetic field only. In the absence of a external magnetic field, the strain yields to an exponential damped amplitude, whereas the curvature leads to a power--law damping of the wave function. The spin--curvature coupling breaks the chiral symmetry between the upper and the lower spinor component, which leads to the increasing of the wave function on either upper or lower region of the wormhole, i.e., depending on the spin number. By adding an uniform magnetic field, the effective potential exhibits an asymptotic quadratic profile and a spin--curvature barrier near the throat. As a result, the bound states (Landau levels) are confined around the wormhole throat showing an asymmetric and spin--dependent profile., Comment: 12 pages, two column, 21 figures

Copper is a good CO2 electroreduction catalyst as products beyond CO form, but efficiency and selectivity is low. Experiments have shown that admixture of other elements can help, and computational screening studies have pointed out various promising candidates based on the adsorption of a single CO molecule as a descriptor. Our calculations of CO adsorption on surfaces where a first row transition metal atom replaces a Cu atom show that multiple CO molecules, not just one, bind to the substitutional atom. For Fe, Co, and Ni atoms, a decrease in binding energy is found, but the reverse trend, namely increasing bond strength, is found for V, Cr, and Mn and the first three CO molecules. Magnetic moment, charge, and position of the substitutional atom are also strongly affected by the CO adsorption in most cases. Magnetic moment is stepwise reduced to zero, and the outward displacement of the substitutional atom increased., Comment: 8 pages, 2 figures

Motivated by the recent experimental developments in van der Waals heterostructures, we investigate the emergent magnetism in Mott insulator - semimetal moir\'e superlattices by deriving effective spin models and exploring their phase diagram by Monte Carlo simulations. Our analysis indicates that the stacking-dependent interlayer Kondo interaction can give rise to different types of magnetic order, forming domains within the moir\'e unit cell. In particular, we find that the AB (AA) stacking regions tend to order (anti)ferromagnetically for an extended range of parameters. The remaining parts of the moir\'e unit cell form ferromagnetic chains that are coupled antiferromagnetically. We show that the decay length of the Kondo interaction can control the extent of these phases. Our results highlight the importance of stacking-dependent interlayer exchange and the rich magnetic spin textures that can be obtained in van der Waals heterostructures.

We update the ephemerides of 16 transiting exoplanets using our ground-based observations, new TESS data, and previously published observations including those of amateur astronomers. All these light curves were modeled by making use of a set of quantitative criteria with the EXOFAST code to obtain mid-transit times. We searched for statistically significant secular and/or periodic trends in the mid-transit times. We found that the timing data are well modeled by a linear ephemeris for all systems except for XO-2 b, for which we detect an orbital decay with the rate of -12.95 $\pm$ 1.85 ms/yr that can be confirmed with future observations. We also detect a hint of potential periodic variations in the TTV data of HAT-P-13 b which also requires confirmation with further precise observations., Comment: 23 pages, 17 figures

In this paper we investigate the behaviour of massless fermions in the black string spacetime by computing the eigenvalues and eigenfunctions of the Weyl equations. These solutions allowed us to study the behaviour of such massless fermions in terms of the cosmological constant, the black string's mass and the radial distance of the particle from the black string. The solutions, written in terms of Parabolic Cylinder functions and Laguerre polynomials, were obtained for a particle far from the black string and around the horizon event. For the particle around the event horizon, for all configuration of parameters, the energy eigenvalues are complex-valued, indicating QNM similarly to the case of spherical black holes. For the particle far from the black string, the energies derived from the Weyl equation set up conditions on the parameters in order to keep the energy as a real valued parameter., Comment: 5 pages, two columns, 1 figure

We present results regarding the longitudinal migrations of cool stellar spots that exhibit remarkable oscillations and explore their possible causes. We conducted analyses using high-quality data from nine target systems of various spectral types, spanning from F to M, which were observed by the Kepler Satellite. The systems in which the behaviour of the spots was examined are as follows: KIC 4357272, KIC 6025466, KIC 6058875, KIC 6962018, KIC 7798259, KIC 9210828, KIC 11706658, KIC 12599700, and KIC 8669092. Basic stellar parameters were calculated from light curve analysis using the PHOEBE V.0.32 software, and light curves were modelled to obtain sinusoidal variations occurring out-of-eclipses phases, induced by rotational modulation. Subsequently, we calculated the minimum times of the obtained sinusoidal variations using the Fourier transform. The distributions of {\theta}min corresponding to these minimum times over time were computed using linear fits to determine the longitudinal migrations of the spotted areas. We then compared the longitudinal migration periods with the stellar parameters found in the literature. In addition, we also found a secondary variation in the spot migrations apart from the linear models. Our results revealed that the longitudinal migration periods vary in relation to the B - V colour index of the stars., Comment: 7 figures, 6 tables, accepted for publication in PASA

The recently proposed combination of the valence-space in-medium similarity renormalization group (VS-IMSRG) with the density matrix renormalization group (DMRG) offers a scalable and flexible many-body approach for strongly correlated open-shell nuclei. We use the VS-DMRG to investigate the low-lying spectroscopy of N=50 isotones, which are characteristic for their transition between single-particle and collective excitations. We also study electromagnetic transitions and show the advantage of the VS-DMRG to capture the underlying physics more efficiently, with significantly improved convergence compared to state-of-the-art shell-model truncations. Combined with an analysis of quantum information measures, this further establishes the VS-DMRG as a valuable method for ab initio calculations of nuclei., Comment: 7 pages, 4 figures, version published at Phys. Lett. B

Topology, like symmetry, is a fundamental concept in understanding general properties of physical systems. In condensed matter, nontrivial topology may manifest itself as singular features in the energy spectrum or the quantization of electrical properties such as conductance and magnetic flux. Using microwave spectroscopy, we determine that a superconducting circuit with three Josephson tunnel junctions in parallel can possess degeneracies indicative of \emph{intrinsic} nontrivial topology. We identify three topological invariants, one of which is related to a hidden quantum mechanical supersymmetry. Measurements show that devices fabricated in different topological regimes fall on a simple phase diagram which should be robust to junction imperfections and geometric inductance. Josephson tunnel junction circuits, which are readily fabricated with conventional microlithography techniques, allow access to a wide range of topological systems that may have no condensed matter analog. Notable spectral features of these circuits, such as degeneracies and flat bands, may find use in quantum information, sensing, and metrology., Comment: 20 pages, 12 figures, published version

Surface acoustic waves (SAWs) hold a vast potential in various fields such as spintronics, quantum acoustics, and electron-quantum optics, but an electromagnetic wave emanating from SAW generation circuits has often been a major hurdle. Here, we investigate a differential excitation method of interdigital transducers (IDTs) to generate SAWs while reducing the electromagnetic wave. The results show that electromagnetic waves are suppressed by more than 90% in all directions. This suppression overcomes the operating limits and improves the scalability of SAW systems. Our results promise to facilitate the development of SAW-based applications in a wide range of research fields.

The structure of the solvation shell of aqueous Fe$^{+3}$ ion has been a subject of controversy due to discrepancies between experiments and different levels of theory. We address this issue by performing simulations for a wide range of ion concentrations, using various empirical potential energy functions, as well as density functional theory calculations of selected configurations. The solvation shell undergoes abrupt transitions between two states: an octahedral (OH) state with 6-fold coordination, and a capped trigonal prism (CTP) state with 7-fold coordination. The lifetime of these states is concentration dependent. In dilute $\mathrm{FeCl_3}$ solutions, the lifetime of the two states is similar ($\approx 1$ ns). However, the lifetime of the OH state increases with ion concentration, while that of the CTP state decreases slightly. When a uniform negative background charge is used instead of explicit counterions, the lifetime of the OH state is greatly overestimated. These findings underscore the need for further experimental measurements as well as high-level simulations over sufficiently long timescales and low concentration.

Surface acoustic waves (SAWs) are a reliable solution to transport single electrons with precision in piezoelectric semiconductor devices. Recently, highly efficient single-electron transport with a strongly compressed single-cycle acoustic pulse has been demonstrated. This approach, however, requires surface gates constituting the quantum dots, their wiring, and multiple gate movements to load and unload the electrons, which is very time-consuming. Here, on the contrary, we employ such a single-cycle acoustic pulse in a much simpler way - without any quantum dot at the entrance or exit of a transport channel - to perform single-electron transport between distant electron reservoirs. We observe the transport of a solitary electron in a single-cycle acoustic pulse via the appearance of the quantized acousto-electric current. The simplicity of our approach allows for on-demand electron emission with arbitrary delays on a ns time scale. We anticipate that enhanced synthesis of the SAWs will facilitate electron-quantum-optics experiments with multiple electron flying qubits.

Large language models (LLMs) have achieved remarkable advancements in natural language understanding and generation. However, one major issue towards their widespread deployment in the real world is that they can generate "hallucinated" answers that are not factual. Towards this end, this paper focuses on improving LLMs by grounding their responses in retrieved passages and by providing citations. We propose a new framework, AGREE, Adaptation for GRounding EnhancEment, that improves the grounding from a holistic perspective. Our framework tunes LLMs to selfground the claims in their responses and provide accurate citations to retrieved documents. This tuning on top of the pre-trained LLMs requires well-grounded responses (with citations) for paired queries, for which we introduce a method that can automatically construct such data from unlabeled queries. The selfgrounding capability of tuned LLMs further grants them a test-time adaptation (TTA) capability that can actively retrieve passages to support the claims that have not been grounded, which iteratively improves the responses of LLMs. Across five datasets and two LLMs, our results show that the proposed tuningbased AGREE framework generates superior grounded responses with more accurate citations compared to prompting-based approaches and post-hoc citing-based approaches, Comment: NAACL 2024

Text-to-SQL aims to automate the process of generating SQL queries on a database from natural language text. In this work, we propose "SQLPrompt", tailored to improve the few-shot prompting capabilities of Text-to-SQL for Large Language Models (LLMs). Our methods include innovative prompt design, execution-based consistency decoding strategy which selects the SQL with the most consistent execution outcome among other SQL proposals, and a method that aims to improve performance by diversifying the SQL proposals during consistency selection with different prompt designs ("MixPrompt") and foundation models ("MixLLMs"). We show that \emph{SQLPrompt} outperforms previous approaches for in-context learning with few labeled data by a large margin, closing the gap with finetuning state-of-the-art with thousands of labeled data.

Estimation of temporal counterfactual outcomes from observed history is crucial for decision-making in many domains such as healthcare and e-commerce, particularly when randomized controlled trials (RCTs) suffer from high cost or impracticality. For real-world datasets, modeling time-dependent confounders is challenging due to complex dynamics, long-range dependencies and both past treatments and covariates affecting the future outcomes. In this paper, we introduce Counterfactual Self-Supervised Transformer (COSTAR), a novel approach that integrates self-supervised learning for improved historical representations. We propose a component-wise contrastive loss tailored for temporal treatment outcome observations and explain its effectiveness from the view of unsupervised domain adaptation. COSTAR yields superior performance in estimation accuracy and generalization to out-of-distribution data compared to existing models, as validated by empirical results on both synthetic and real-world datasets.

In this paper we consider the finite element approximation of Maxwell's problem and analyse the prescription of essential boundary conditions in a weak sense using Nitsche's method. To avoid indefiniteness of the problem, the original equations are augmented with the gradient of a scalar field that allows one to impose the zero divergence of the magnetic induction, even if the exact solution for this scalar field is zero. Two finite element approximations are considered, namely, one in which the approximation spaces are assumed to satisfy the appropriate inf-sup condition that render the standard Galerkin method stable, and another augmented and stabilised one that permits the use of finite element interpolations of arbitrary order. Stability and convergence results are provided for the two finite element formulations considered.

We develop an exactly solvable model with Kitaev-type interactions and study its phase diagram on the dual lattice of the quasicrystalline Ammann-Beenker lattice. Our construction is based on the $\Gamma$-matrix generalization of the Kitaev model and utilizes the cut-and-project correspondence between the four-dimensional simple cubic lattice and the Ammann-Beenker lattice to designate four types of bonds. We obtain a rich phase diagram with gapped (chiral and abelian) and gapless spin liquid phases via Monte Carlo simulations and variational analysis. We show that the ground state can be further tuned by the inclusion of an onsite term that selects 21 different vison configurations while maintaining the integrability of the model. Our results highlight the rich physics at the intersection of quasicrystals and quantum magnetism.

Bose mixture quantum droplets display a fascinating stability that relies on quantum fluctuations to prevent collapse driven by mean-field effects. Most droplet research focuses on untrapped or weakly trapped scenarios, where the droplets exhibit a liquid-like flat density profile. When weakly trapped droplets rotate, they usually respond through center-of-mass motion or splitting instability. Here, we study rapidly rotating droplets in the strong external confinement limit where the external potential prevents splitting and the center-of-mass excitation. We find that quantum droplets form a triangular vortex lattice as in single-component repulsive Bose-Einstein condensates (BEC), but the overall density follows the analytical Thomas-Fermi profile obtained from a cubic equation. We observe three significant differences between rapidly rotating droplets and repulsive BECs. First, the vortex core size changes markedly at finite density, visible in numerically obtained density profiles. We analytically estimate the vortex core sizes from the droplets' coherence length and find good agreement with the numerical results. Second, the change in the density profile gives a slight but observable distortion to the lattice, which agrees with the distortion expected due to nonuniform superfluid density. Lastly, unlike a repulsive BEC, which expands substantially as the rotation frequency approaches the trapping frequency, rapidly rotating droplets show only a fractional change in their size. We argue that this last point can be used to create clouds with lower filling factors, which may facilitate reaching the elusive strongly correlated regime.

We report the first measurement of CNO solar neutrinos by Borexino that uses the Correlated Integrated Directionality (CID) method, exploiting the sub-dominant Cherenkov light in the liquid scintillator detector. The directional information of the solar origin of the neutrinos is preserved by the fast Cherenkov photons from the neutrino scattered electrons, and is used to discriminate between signal and background. The directional information is independent from the spectral information on which the previous CNO solar neutrino measurements by Borexino were based. While the CNO spectral analysis could only be applied on the Phase-III dataset, the directional analysis can use the complete Borexino data taking period from 2007 to 2021. The absence of CNO neutrinos has been rejected with >5{\sigma} credible level using the Bayesian statistics. The directional CNO measurement is obtained without an external constraint on the $^{210}$Bi contamination of the liquid scintillator, which was applied in the spectral analysis approach. The final and the most precise CNO measurement of Borexino is then obtained by combining the new CID-based CNO result with an improved spectral fit of the Phase-III dataset. Including the statistical and the systematic errors, the extracted CNO interaction rate is $R(\mathrm{CNO})=6.7^{+1.2}_{-0.8} \, \mathrm{cpd/100 \, tonnes}$. Taking into account the neutrino flavor conversion, the resulting CNO neutrino flux at Earth is $\Phi_\mathrm{CNO}=6.7 ^{+1.2}_{-0.8} \times 10^8 \, \mathrm{cm^{-2} s^{-1}}$, in agreement with the high metallicity Standard Solar Models. The results described in this work reinforce the role of the event directional information in large-scale liquid scintillator detectors and open up new avenues for the next-generation liquid scintillator or hybrid neutrino experiments.

Text-to-SQL, the process of translating natural language into Structured Query Language (SQL), represents a transformative application of large language models (LLMs), potentially revolutionizing how humans interact with data. This paper introduces the SQL-PaLM framework, a comprehensive solution for understanding and enhancing Text-to-SQL using LLMs, using in the learning regimes of few-shot prompting and instruction fine-tuning. With few-shot prompting, we explore the effectiveness of consistency decoding with execution-based error filtering. With instruction fine-tuning, we delve deep in understanding the critical paradigms that influence the performance of tuned LLMs. In particular, we investigate how performance can be improved through expanded training data coverage and diversity, synthetic data augmentation, and integrating query-specific database content. We propose a test-time selection method to further refine accuracy by integrating SQL outputs from multiple paradigms with execution feedback as guidance. Additionally, we tackle the practical challenge of navigating intricate databases with a significant number of tables and columns, proposing efficient techniques for accurately selecting relevant database elements to enhance Text-to-SQL performance. Our holistic approach yields substantial advancements in Text-to-SQL, as demonstrated on two key public benchmarks, Spider and BIRD. Through comprehensive ablations and error analyses, we shed light on the strengths and weaknesses of our framework, offering valuable insights into Text-to-SQL's future work.

We develop a class of interacting particle systems for implementing a maximum marginal likelihood estimation (MMLE) procedure to estimate the parameters of a latent variable model. We achieve this by formulating a continuous-time interacting particle system which can be seen as a Langevin diffusion over an extended state space of parameters and latent variables. In particular, we prove that the parameter marginal of the stationary measure of this diffusion has the form of a Gibbs measure where number of particles acts as the inverse temperature parameter in classical settings for global optimisation. Using a particular rescaling, we then prove geometric ergodicity of this system and bound the discretisation error in a manner that is uniform in time and does not increase with the number of particles. The discretisation results in an algorithm, termed Interacting Particle Langevin Algorithm (IPLA) which can be used for MMLE. We further prove nonasymptotic bounds for the optimisation error of our estimator in terms of key parameters of the problem, and also extend this result to the case of stochastic gradients covering practical scenarios. We provide numerical experiments to illustrate the empirical behaviour of our algorithm in the context of logistic regression with verifiable assumptions. Our setting provides a straightforward way to implement a diffusion-based optimisation routine compared to more classical approaches such as the Expectation Maximisation (EM) algorithm, and allows for especially explicit nonasymptotic bounds., Comment: 38 pages