Search

We establish a generalized quantum asymptotic equipartition property (AEP) beyond the i.i.d. framework where the random samples are drawn from two sets of quantum states. In particular, under suitable assumptions on the sets, we prove that all operationally relevant divergences converge to the quantum relative entropy between the sets. More specifically, both the smoothed min- and max-relative entropy approach the regularized relative entropy between the sets. Notably, the asymptotic limit has explicit convergence guarantees and can be efficiently estimated through convex optimization programs, despite the regularization, provided that the sets have efficient descriptions. We give four applications of this result: (i) The generalized AEP directly implies a new generalized quantum Stein's lemma for conducting quantum hypothesis testing between two sets of quantum states. (ii) We introduce a quantum version of adversarial hypothesis testing where the tester plays against an adversary who possesses internal quantum memory and controls the quantum device and show that the optimal error exponent is precisely characterized by a new notion of quantum channel divergence, named the minimum output channel divergence. (iii) We derive a relative entropy accumulation theorem stating that the smoothed min-relative entropy between two sequential processes of quantum channels can be lower bounded by the sum of the regularized minimum output channel divergences. (iv) We apply our generalized AEP to quantum resource theories and provide improved and efficient bounds for entanglement distillation, magic state distillation, and the entanglement cost of quantum states and channels. At a technical level, we establish new additivity and chain rule properties for the measured relative entropy which we expect will have more applications., Comment: comments are welcome

Superluminal propagation is an intrinsic problem in the diffusion equation and has not been effectively addressed for a long time. In this work, a rigorous solution to this issue is obtained under the assumption that particles undergo a random flight process, where they move isotropically at a constant speed while experiencing random scatterings. We validate this solution by comparing it with comprehensive simulations of the random flight process and find that it significantly deviates from the solution derived from the J\"{u}ttner propagator. This solution is broadly applicable to various diffusion phenomena, such as cosmic-ray propagation. We emphasize that our rigorous solution is particularly crucial in scenarios involving burst-like particle injection, where previous phenomenological approaches to the superluminal diffusion problem may not yield accurate results., Comment: 8 pages, 4 figures

We introduce Baichuan Alignment, a detailed analysis of the alignment techniques employed in the Baichuan series of models. This represents the industry's first comprehensive account of alignment methodologies, offering valuable insights for advancing AI research. We investigate the critical components that enhance model performance during the alignment process, including optimization methods, data strategies, capability enhancements, and evaluation processes. The process spans three key stages: Prompt Augmentation System(PAS), Supervised Fine-Tuning(SFT), and Preference Alignment. The problems encountered, the solutions applied, and the improvements made are thoroughly recorded. Through comparisons across well-established benchmarks, we highlight the technological advancements enabled by Baichuan Alignment. Baichuan-Instruct is an internal model, while Qwen2-Nova-72B and Llama3-PBM-Nova-70B are instruct versions of the Qwen2-72B and Llama-3-70B base models, optimized through Baichuan Alignment. Baichuan-Instruct demonstrates significant improvements in core capabilities, with user experience gains ranging from 17% to 28%, and performs exceptionally well on specialized benchmarks. In open-source benchmark evaluations, both Qwen2-Nova-72B and Llama3-PBM-Nova-70B consistently outperform their respective official instruct versions across nearly all datasets. This report aims to clarify the key technologies behind the alignment process, fostering a deeper understanding within the community. Llama3-PBM-Nova-70B model is available at https://huggingface.co/PKU-Baichuan-MLSystemLab/Llama3-PBM-Nova-70B.

Quantum resource distillation is a fundamental task in quantum information science. Minimizing the distillation overhead, i.e., the amount of noisy source states required to produce some desired output state within some target error, is crucial for the scalability of quantum computation and communication. Here, we show that quantum catalysts -- an additional resource that facilitates the transformation but remains unchanged before and after the process -- can help surpass previously known fundamental limitations on distillation overhead. Specifically, we show that multi-shot distillation protocols can be converted into one-shot catalytic protocols, which hold significant practical benefits, while maintaining the distillation overhead. In particular, in the context of magic state distillation, our result indicates that the code-based low-overhead distillation protocols that rely on divergingly large batches can be promoted to the one-shot setting where the batch volume can be arbitrarily small for any accuracy. Combining with very recent results on asymptotically good quantum codes with transversal non-Clifford gates, we demonstrate that magic state distillation with constant overhead can be achieved with controllable output size using catalytic protocols. Furthermore, we demonstrate that catalysis enables a spacetime trade-off between overhead and success probability. Notably, we show that the optimal constant for constant-overhead catalytic magic state distillation can be reduced to $1$ at the price of compromising the success probability by a constant factor. Finally, we present an illustrative example that extends the catalysis techniques to the study of dynamic quantum resources. This provides the channel mutual information with a one-shot operational interpretation, thereby addressing an open question posed by Wilming., Comment: 13 pages, 3 figures; comments are welcome

Out-of-Distribution (OoD) detection aims to justify whether a given sample is from the training distribution of the classifier-under-protection, i.e., In-Distribution (InD), or from OoD. Diffusion Models (DMs) are recently utilized in OoD detection by using the perceptual distances between the given image and its DM generation. DM-based methods bring fresh insights to the field, yet remain under-explored. In this work, we point out two main limitations in DM-based OoD detection methods: (i) the perceptual metrics on the disparities between the given sample and its generation are devised only at human-perceived levels, ignoring the abstract or high-level patterns that help better reflect the intrinsic disparities in distribution; (ii) only the raw image contents are taken to measure the disparities, while other representations, i.e., the features and probabilities from the classifier-under-protection, are easy to access at hand but are ignored. To this end, our proposed detection framework goes beyond the perceptual distances and looks into the deep representations from the classifier-under-protection with our novel metrics devised correspondingly, leading to more informative disparity assessments between InD and OoD. An anomaly-removal strategy is integrated to remove the abnormal OoD information in the generation, further enhancing the distinctiveness of disparities. Our work has demonstrated state-of-the-art detection performances among DM-based methods in extensive experiments.

The recent AMS-02 measurements of cosmic-ray (CR) deuteron fluxes suggest the presence of primary deuterons in quantities far exceeding predictions from Big Bang nucleosynthesis. This poses a significant challenge to modern astrophysics, as no known processes can account for such large amounts of deuterons without violating existing constraints~\cite{Epstein:1976hq}. In contrast, it is recently proposed that the AMS-02 measurements can be explained by a purely secondary origin when contributions from heavier nuclei are considered. In this study, we recalculate the secondary deuteron flux using production cross sections updated with the latest collider data. We find that some of the deuteron production cross sections are overestimated in the widely-used calculation tools for CR propagation, and a primary deuteron component is still necessary. We then propose a novel process for generating primary deuterons at CR sources through a fusion mechanism, which is naturally unique to deuterons. This model could explain the observed deuteron excess while maintaining consistency with other CR measurements., Comment: 8 pages, 8 figures

The gamma-ray halo surrounding Geminga suggests a notable reduction in cosmic-ray diffusion. One potential explanation for this phenomenon is the projection effect of slow diffusion perpendicular to the average magnetic field (represented by the diffusion coefficient $D_\perp$) within an anisotropic diffusion framework. In this context, the diffusion coefficient parallel to the mean field ($D_\parallel$) may remain substantial, allowing electrons and positrons ($e^\pm$) generated by Geminga to effectively propagate towards Earth along magnetic field lines, potentially leading to an observable $e^\pm$ flux. This study initially establishes the fundamental parameters of the anisotropic model based on the morphology and spectral observations of the Geminga halo, and subsequently forecasts the $e^\pm$ flux generated by Geminga at Earth's location. Our findings indicate that the $e^-+e^+$ spectrum obtained by DAMPE can provide critical constraints on the anisotropic diffusion model: to ensure that the projected spectrum does not surpass the observational data, the Alfv\'en Mach number of the turbulent magnetic field ($M_A$) should not fall below 0.75, corresponding to $D_\parallel/D_\perp\lesssim3$ given $D_\perp=D_\parallel M_A^4$. This suggests that a substantial reduction in $D_\parallel$ relative to the Galactic average may still be necessary. Additionally, our analysis reveals that within the anisotropic diffusion framework, Geminga could generate a distinct peak around 1 TeV in the $e^-+e^+$ spectrum, potentially accounting for the anomalous 1.4 TeV excess tentatively detected by DAMPE., Comment: 8 pages,6 figures

The Large High-Altitude Air Shower Observatory (LHAASO) recently published measurements of diffuse Galactic gamma-ray emission (DGE) in the 10-1000 TeV energy range. The measured DGE flux is significantly higher than the expectation from hadronic interactions between cosmic rays (CRs) and the interstellar medium. This excess has been proposed to originate from unknown extended sources produced by electron radiation, such as pulsar wind nebulae or pulsar halos (PWNe/halos). In this study, we propose a new perspective to explain the DGE excess observed by LHAASO. The masking regions used in the LHAASO DGE measurement may not fully encompass the extended signals of PWNe/halos. By employing a two-zone diffusion model for electrons around pulsars, we find that the DGE excess in most regions of the Galactic plane can be well explained by the signal leakage model under certain parameters. Our results indicate that the signal leakage from known sources and contributions from unresolved sources should be considered complementary in explaining the DGE excess., Comment: 16 pages, 5 figures, to be published in Chinese Physics C

In recent years, the rise of Large Language Models (LLMs) has spurred a growing demand for plug-and-play AI systems. Among the various AI techniques, prompt engineering stands out as particularly significant. However, users often face challenges in writing prompts due to the steep learning curve and significant time investment, and existing automatic prompt engineering (APE) models can be difficult to use. To address this issue, we propose PAS, an LLM-based plug-and-play APE system. PAS utilizes LLMs trained on high-quality, automatically generated prompt complementary datasets, resulting in exceptional performance. In comprehensive benchmarks, PAS achieves state-of-the-art (SoTA) results compared to previous APE models, with an average improvement of 6.09 points. Moreover, PAS is highly efficient, achieving SoTA performance with only 9000 data points. Additionally, PAS can autonomously generate prompt augmentation data without requiring additional human labor. Its flexibility also allows it to be compatible with all existing LLMs and applicable to a wide range of tasks. PAS excels in human evaluations, underscoring its suitability as a plug-in for users. This combination of high performance, efficiency, and flexibility makes PAS a valuable system for enhancing the usability and effectiveness of LLMs through improved prompt engineering.

Retrieval-Augmented Generation (RAG) offers a promising solution to address various limitations of Large Language Models (LLMs), such as hallucination and difficulties in keeping up with real-time updates. This approach is particularly critical in expert and domain-specific applications where LLMs struggle to cover expert knowledge. Therefore, evaluating RAG models in such scenarios is crucial, yet current studies often rely on general knowledge sources like Wikipedia to assess the models' abilities in solving common-sense problems. In this paper, we evaluated LLMs by RAG settings in a domain-specific context, college enrollment. We identified six required abilities for RAG models, including the ability in conversational RAG, analyzing structural information, faithfulness to external knowledge, denoising, solving time-sensitive problems, and understanding multi-document interactions. Each ability has an associated dataset with shared corpora to evaluate the RAG models' performance. We evaluated popular LLMs such as Llama, Baichuan, ChatGLM, and GPT models. Experimental results indicate that existing closed-book LLMs struggle with domain-specific questions, highlighting the need for RAG models to solve expert problems. Moreover, there is room for RAG models to improve their abilities in comprehending conversational history, analyzing structural information, denoising, processing multi-document interactions, and faithfulness in expert knowledge. We expect future studies could solve these problems better.

Improving the generalization ability of modern deep neural networks (DNNs) is a fundamental challenge in machine learning. Two branches of methods have been proposed to seek flat minima and improve generalization: one led by sharpness-aware minimization (SAM) minimizes the worst-case neighborhood loss through adversarial weight perturbation (AWP), and the other minimizes the expected Bayes objective with random weight perturbation (RWP). While RWP offers advantages in computation and is closely linked to AWP on a mathematical basis, its empirical performance has consistently lagged behind that of AWP. In this paper, we revisit the use of RWP for improving generalization and propose improvements from two perspectives: i) the trade-off between generalization and convergence and ii) the random perturbation generation. Through extensive experimental evaluations, we demonstrate that our enhanced RWP methods achieve greater efficiency in enhancing generalization, particularly in large-scale problems, while also offering comparable or even superior performance to SAM. The code is released at https://github.com/nblt/mARWP., Comment: Accepted to TMLR 2024

Recently LHAASO Collaboration gives precise measurements of cosmic rays (CR) all particle energy spectrum and mean logarithmic mass $\left \langle \ln A \right \rangle$ from 0.3 PeV to 30 PeV. Combining the CR measurements by AMS-02 and DAMPE in space and that by LHAASO and Auger on the ground we construct a model to recover all these measurements from tens of GeV to tens of EeV. We find the LHAASO measurement is crucial in the model construction by connecting the Galactic component to the extragalactic component. The precise measurements of CR spectra for individual species by AMS-02 and DAMPE together with the newest LHAASO results clearly indicates three Galactic CR components, that is, a soft low energy background, a hard high energy component, and a local source contribution. However, the LHAASO data show that above $\sim 10^{16}$ eV a nonnegligible extragalactic component must be included. Combining the Auger results and the LHAASO results we figure out the extragalactic CRs which need at least two components at lower and higher energies. Thanks to the precise measurements by LHAASO the constraints on the model parameters are quite stringent. The spectra features and mass measurements in all energy range are all well reproduced in the model., Comment: 11 pages, 2 figures, 4 tables

The excesses in the electron and positron spectra observed by many experiments, such as PAMELA and AMS-02, have sparked significant theoretical investigation. It is not easy to distinguish the two primary hypotheses dark matter annihilation/decay and pulsars from the spectral features. Should pulsars be the source of this excess, the expected variability in their distribution may introduce distinct irregularities in the positron energy spectrum. In this study, we use an irregularity estimator to detect these potential features in the positron energy spectrum of AMS-02. Our analysis of the current AMS-02 data reveals these spectral irregularities with a statistical significance of $1.75\sigma$. However, our projection indicates that, with AMS-02 data collected over a period of 20 years, such irregularities could be identified with a confidence level of $3\sigma$ level in 71\% of our simulations., Comment: 6 pages, 6 figures

The integration of large language models (LLMs) and search engines represents a significant evolution in knowledge acquisition methodologies. However, determining the knowledge that an LLM already possesses and the knowledge that requires the help of a search engine remains an unresolved issue. Most existing methods solve this problem through the results of preliminary answers or reasoning done by the LLM itself, but this incurs excessively high computational costs. This paper introduces a novel collaborative approach, namely SlimPLM, that detects missing knowledge in LLMs with a slim proxy model, to enhance the LLM's knowledge acquisition process. We employ a proxy model which has far fewer parameters, and take its answers as heuristic answers. Heuristic answers are then utilized to predict the knowledge required to answer the user question, as well as the known and unknown knowledge within the LLM. We only conduct retrieval for the missing knowledge in questions that the LLM does not know. Extensive experimental results on five datasets with two LLMs demonstrate a notable improvement in the end-to-end performance of LLMs in question-answering tasks, achieving or surpassing current state-of-the-art models with lower LLM inference costs., Comment: Accepted by ACL 2024 main conference. Repo: https://github.com/plageon/SlimPLM

Previous studies on the fitting of spectral energy distributions (SEDs) often apply the external-Compton process to interpret the high-energy peak of low-synchrotron-peaked (LSP) BL Lac objects (LBLs), despite the lack of strong broad emission lines observed for LBLs. In this work, we collect quasi-simultaneous multi-wavelength data of 15 LBLs from the Fermi fourth LAT AGN catalog (4LAC). We propose an analytical method to assess the necessity of external photon fields in the framework of one-zone scenario. Following derived analytical results, we fit the SEDs of these LBLs with the conventional one-zone leptonic model and study their jet physical properties. Our main results can be summarized as follows. (1)We find that most LBLs cannot be fitted by the one-zone synchrotron self-Compton (SSC) model. This indicates that external photons play a crucial role in the high-energy emission of LBLs, therefore we suggest that LBLs are masquerading BL Lacs. (2) We suggest that the $\gamma$-ray emitting regions of LBLs are located outside the broad-line region and within the dusty torus. (3) By extending the analytical method to all types of LSPs in Fermi-4LAC (using historical data), we find that the high-energy peaks of some flat spectrum radio quasars and blazar candidates of unknown types can be attributed to the SSC emission, implying that the importance of external photons could be minor. We suggest that the variability timescale may help distinguish the origin of the high-energy peak., Comment: Accepted for publication in MNRAS (14 pages, 7 figures, 2 tables)

The Be isotopic measurements preliminarily reported by the AMS-02 Collaboration have reached an unprecedented energy of 12 GeV/$n$. As secondary cosmic rays (CRs), the Be isotopes include both stable and unstable species, which are crucial for constraining the propagation parameters of Galactic CRs. However, uncertainties in their production cross sections can skew the interpretation of the CR data, especially when cross-section measurements are of significantly lower quality than CR measurements. In this work, we consider the uncertainties of the cross sections to interpret the Be isotopic data by adopting a cross-section parametrization that fully utilizes the available experimental data. Owing to the high-quality measurements of the $^7$Be production cross section, we innovatively employ $^7$Be instead of $^9$Be to constrain propagation parameters. Notably, the diffusion halo thickness is constrained to $5.67\pm0.76$~kpc, representing a moderate value compared to previous analogous works. Combining the well-constrained CR propagation model and the precise CR measurements of $^9$Be, we conversely constrain the major production cross section of $^9$Be and find that it ought to be remarkably lower than previously thought. Our analysis also questions the reliability of certain cross sections measured by some experiments, potentially marking the first time CR data has been used to identify dubious nucleon production cross sections. The method presented in this work holds promise for analyzing upcoming isotopic data from other nuclei., Comment: 15 pages, 9 figures

Out-of-Distribution (OoD) detection is vital for the reliability of Deep Neural Networks (DNNs). Existing works have shown the insufficiency of Principal Component Analysis (PCA) straightforwardly applied on the features of DNNs in detecting OoD data from In-Distribution (InD) data. The failure of PCA suggests that the network features residing in OoD and InD are not well separated by simply proceeding in a linear subspace, which instead can be resolved through proper non-linear mappings. In this work, we leverage the framework of Kernel PCA (KPCA) for OoD detection, and seek suitable non-linear kernels that advocate the separability between InD and OoD data in the subspace spanned by the principal components. Besides, explicit feature mappings induced from the devoted task-specific kernels are adopted so that the KPCA reconstruction error for new test samples can be efficiently obtained with large-scale data. Extensive theoretical and empirical results on multiple OoD data sets and network structures verify the superiority of our KPCA detector in efficiency and efficacy with state-of-the-art detection performance., Comment: Accepted by NeurIPS 2024

We study entanglement distillation and dilution of states and channels in the single-shot regime. With the help of a recently introduced conversion distance, we provide compact closed-form expressions for the dilution and distillation of pure states and show how this can be used to efficiently calculate these quantities on multiple copies of pure states. These closed-form expressions also allow us to obtain second-order asymptotics. We then prove that the epsilon-single-shot entanglement cost of mixed states is given exactly in terms of an expression containing a suitably smoothed version of the conditional max-entropy. For pure states, this expression reduces to the smoothed max-entropy of the reduced state. Based on these results, we bound the single-shot entanglement cost of channels. We then turn to the one-way entanglement distillation of states and channels and provide bounds in terms of a quantity we denote coherent information of entanglement., Comment: 10+33 pages, comments welcome!

Quantum circuit mapping is a crucial process in the quantum circuit compilation pipeline, facilitating the transformation of a logical quantum circuit into a list of instructions directly executable on a target quantum system. Recent research has introduced a post-compilation step known as remapping, which seeks to reconfigure the initial circuit mapping to mitigate quantum circuit errors arising from system variability. As quantum processors continue to scale in size, the efficiency of quantum circuit mapping and the overall compilation process has become of paramount importance. In this work, we introduce a quantum circuit remapping algorithm that leverages the intrinsic symmetries in quantum processors, making it well-suited for large-scale quantum systems. This algorithm identifies all topologically equivalent circuit mappings by constraining the search space using symmetries and accelerates the scoring of each mapping using vector computation. Notably, this symmetry-based circuit remapping algorithm exhibits linear scaling with the number of qubits in the target quantum hardware and is proven to be optimal in terms of its time complexity. Moreover, we conduct a comparative analysis against existing methods in the literature, demonstrating the superior performance of our symmetry-based method on state-of-the-art quantum hardware architectures and highlighting the practical utility of our algorithm, particularly for quantum processors with millions of qubits., Comment: 10 pages, 5 figures; comments are welcome

Measuring the electron diffusion coefficient is the most straightforward task in the study of gamma-ray pulsar halos. The updated measurements of the spatial morphology and spectrum of the Geminga halo by the High-Altitude Water Cherenkov (HAWC) experiment enable us to constrain parameters beyond the diffusion coefficient, including the size of the slow-diffusion zone and the electron injection spectrum from the pulsar wind nebulae (PWNe). Based on the two-zone diffusion model, we find that the slow-diffusion zone size ($r_*$) around Geminga is within the range of $30-70$~pc. The lower boundary of this range is determined by the goodness of fit of the model to the one-dimensional morphology of the Geminga halo. The upper limit is derived from fitting the gamma-ray spectrum of the Geminga halo, along with the expectations for the power-law index of the injection spectrum based on simulations and PWNe observations, i.e., $p\gtrsim1$. With $r_*$ set at its lower limit of $30$~pc, we obtain the maximum $p$ permitted by the HAWC spectrum measurement, with an upper limit of $2.17$ at a $3\sigma$ significance. Moreover, we find that when $r_*=30$~pc and $p=2.17$, the predicted positron spectrum generated by Geminga at Earth coincides with the AMS-02 measurement in the $50-500$~GeV range., Comment: 15 pages, 5 figures. Accepted for publication in PRD

Existing Out-of-Distribution (OoD) detection methods address to detect OoD samples from In-Distribution (InD) data mainly by exploring differences in features, logits and gradients in Deep Neural Networks (DNNs). We in this work propose a new perspective upon loss landscape and mode ensemble to investigate OoD detection. In the optimization of DNNs, there exist many local optima in the parameter space, or namely modes. Interestingly, we observe that these independent modes, which all reach low-loss regions with InD data (training and test data), yet yield significantly different loss landscapes with OoD data. Such an observation provides a novel view to investigate the OoD detection from the loss landscape, and further suggests significantly fluctuating OoD detection performance across these modes. For instance, FPR values of the RankFeat method can range from 46.58% to 84.70% among 5 modes, showing uncertain detection performance evaluations across independent modes. Motivated by such diversities on OoD loss landscape across modes, we revisit the deep ensemble method for OoD detection through mode ensemble, leading to improved performance and benefiting the OoD detector with reduced variances. Extensive experiments covering varied OoD detectors and network structures illustrate high variances across modes and validate the superiority of mode ensemble in boosting OoD detection. We hope this work could attract attention in the view of independent modes in the loss landscape of OoD data and more reliable evaluations on OoD detectors., Comment: published in International Journal of Computer Vision

Quantum computing has shown tremendous promise in addressing complex computational problems, yet its practical realization is hindered by the limited availability of qubits for computation. Recent advancements in quantum hardware have introduced mid-circuit measurements and resets, enabling the reuse of measured qubits and significantly reducing the qubit requirements for executing quantum algorithms. In this work, we present a systematic study of dynamic quantum circuit compilation, a process that transforms static quantum circuits into their dynamic equivalents with a reduced qubit count through qubit-reuse. We establish the first general framework for optimizing the dynamic circuit compilation via graph manipulation. In particular, we completely characterize the optimal quantum circuit compilation using binary integer programming, provide efficient algorithms for determining whether a given quantum circuit can be reduced to a smaller circuit and present heuristic algorithms for devising dynamic compilation schemes in general. Furthermore, we conduct a thorough analysis of quantum circuits with practical relevance, offering optimal compilations for well-known quantum algorithms in quantum computation, ansatz circuits utilized in quantum machine learning, and measurement-based quantum computation crucial for quantum networking. We also perform a comparative analysis against state-of-the-art approaches, demonstrating the superior performance of our methods in both structured and random quantum circuits. Our framework lays a rigorous foundation for comprehending dynamic quantum circuit compilation via qubit-reuse, bridging the gap between theoretical quantum algorithms and their physical implementation on quantum computers with limited resources., Comment: 51 pages, 32 figures; comments are welcome; v2 reorganize the writing and strengthen the results

Large language models (LLMs) have demonstrated remarkable performance on a variety of natural language tasks based on just a few examples of natural language instructions, reducing the need for extensive feature engineering. However, most powerful LLMs are closed-source or limited in their capability for languages other than English. In this technical report, we present Baichuan 2, a series of large-scale multilingual language models containing 7 billion and 13 billion parameters, trained from scratch, on 2.6 trillion tokens. Baichuan 2 matches or outperforms other open-source models of similar size on public benchmarks like MMLU, CMMLU, GSM8K, and HumanEval. Furthermore, Baichuan 2 excels in vertical domains such as medicine and law. We will release all pre-training model checkpoints to benefit the research community in better understanding the training dynamics of Baichuan 2., Comment: Baichuan 2 technical report. Github: https://github.com/baichuan-inc/Baichuan2

Recently, the Large High Altitude Air Shower Observatory (LHAASO) collaboration presented the first catalog of $\gamma$-ray sources using 508 days of LHAASO data, from March 2021 to September 2022. This catalog contains four blazars and a possible liner-type AGN counterpart. In this work, we establish averaged multi-wavelength SEDs by combining data from the \textsl{Fermi}-Large Area Telescope, \textsl{Swift}, ZTF, and WISE covering the same period as the LHAASO detection. In general, these five AGNs are found in low states at all wavelengths. To study the multi-wavelength properties of these AGNs, several jet emission models, including the one-zone leptonic model, the one-zone leptonic and hadronuclear ($pp$) model, the one-zone proton-synchrotron model, and the spine-layer model are applied to reproduce their averaged SEDs, respectively. We find that the one-zone leptonic model can reproduce most of the SEDs, except for the high-energy tail of the LHAASO spectra of Mrk 421 and Mrk 501. To improve the fitting, emission from $pp$ interactions is favoured in the framework of a one-zone model. The spine-layer model, which can be treated as a multi-zone scenario, can also provide good spectral fits. The influence of different extragalactic background light models on fitting LHAASO energy spectrum is also discussed., Comment: 26 pages, 10 figures, accepted by ApJS. Suggestions, comments or feedback will be appreciated!

The positron excess in cosmic rays has stimulated a lot of interests in the last decade. The dark matter origin of the extra positrons has attracted great attention. However, the $\gamma$-ray search set very stringent constraints on the dark matter annihilation/decay rate, which leads to great disfavor of the dark matter scenario. In the work, we incorporate the recent progress in cosmic rays propagation and reexamine the dark matter scenario accounting for the positron excess. Recent observations indicate that cosmic rays propagation in the Milky Way may be not uniform and diffusion in the Galactic disk should be slower than that in the halo. In the spatial-dependent propagation model, the positrons/electrons are more concentrated in the disk and lead to smaller dark matter annihilation/decay rate to account for the positron excess and also a smaller deficit in the background positron flux. Especially for the $\mu^+\mu^-$ channel the positron spectrum fit the AMS-02 latest data perfectly and the annihilation rate satisfies all the present constraints from $\gamma$-ray and CMB observations., Comment: 11 pages, 4 figures

Anisotropic diffusion is one of the potential interpretations for the morphology of the Geminga pulsar halo. It interprets the observed slow-diffusion phenomenon through a geometric effect, assuming the mean magnetic field direction around Geminga is closely aligned with the line of sight toward it. However, this direction should not extend further than the correlation length of the turbulent magnetic field $L_c$, which could be $100$ pc or less. We first revisit the $L_c=\infty$ scenario and show that the halo asymmetry predicted by this scenario is mainly contributed by the electrons located beyond the ``core" section around Geminga, which has a length of $100$ pc. Then, considering the directional variation of the magnetic field beyond the core section, we take one magnetic field configuration as an example to investigate the possible halo morphology. The predicted morphology has some different features compared to the $L_c=\infty$ scenario. The current experiments may already be able to test these features. In addition, we use a semi-analytical method to solve the anisotropic propagation equation, which offers significant convenience compared to numerical approaches., Comment: 15 pages, 7 figures

Recent studies on cosmic rays (CRs) have reported the possibility of an excess in the antiproton flux around $10-20$ GeV. However, the associated systematic uncertainties have impeded the interpretation of these findings. In this study, we conduct a global Bayesian analysis to constrain the propagation parameters and evaluate the CR antiproton spectrum, while comprehensively accounting for uncertainties associated with interstellar CR propagation, production cross sections for antiprotons and other secondaries, and the charge and energy dependent effects of solar modulation. We establish that the most recent AMS-02 $\bar{p}$ spectrum is in agreement with a pure secondary origin. Based on this, we establish upper limits on dark matter (DM) annihilation. We also determine that the AMS-02 data favors the empirical hadronic interaction models over phenomenological ones. Finally, we find that the latest AMS-02 antiproton data from 2011 to 2018 disfavors the antiproton excess at $\mathcal{O}$(10) GeV and the corresponding DM interpretation that can simultaneously account for the Galactic Center excess in the gamma-ray observation., Comment: 11 pages, 7 figures

The quantum internet is envisioned as the ultimate stage of the quantum revolution, which surpasses its classical counterpart in various aspects, such as the efficiency of data transmission, the security of network services, and the capability of information processing. Given its disruptive impact on the national security and the digital economy, a global race to build scalable quantum networks has already begun. With the joint effort of national governments, industrial participants and research institutes, the development of quantum networks has advanced rapidly in recent years, bringing the first primitive quantum networks within reach. In this work, we aim to provide an up-to-date review of the field of quantum networks from both theoretical and experimental perspectives, contributing to a better understanding of the building blocks required for the establishment of a global quantum internet. We also introduce a newly developed quantum network toolkit to facilitate the exploration and evaluation of innovative ideas. Particularly, it provides dual quantum computing engines, supporting simulations in both the quantum circuit and measurement-based models. It also includes a compilation scheme for mapping quantum network protocols onto quantum circuits, enabling their emulations on real-world quantum hardware devices. We showcase the power of this toolkit with several featured demonstrations, including a simulation of the Micius quantum satellite experiment, a testing of a four-layer quantum network architecture with resource management, and a quantum emulation of the CHSH game. We hope this work can give a better understanding of the state-of-the-art development of quantum networks and provide the necessary tools to make further contributions along the way., Comment: 36 pages, 33 figures; comments are welcome

To fully uncover the great potential of deep neural networks (DNNs), various learning algorithms have been developed to improve the model's generalization ability. Recently, sharpness-aware minimization (SAM) establishes a generic scheme for generalization improvements by minimizing the sharpness measure within a small neighborhood and achieves state-of-the-art performance. However, SAM requires two consecutive gradient evaluations for solving the min-max problem and inevitably doubles the training time. In this paper, we resort to filter-wise random weight perturbations (RWP) to decouple the nested gradients in SAM. Different from the small adversarial perturbations in SAM, RWP is softer and allows a much larger magnitude of perturbations. Specifically, we jointly optimize the loss function with random perturbations and the original loss function: the former guides the network towards a wider flat region while the latter helps recover the necessary local information. These two loss terms are complementary to each other and mutually independent. Hence, the corresponding gradients can be efficiently computed in parallel, enabling nearly the same training speed as regular training. As a result, we achieve very competitive performance on CIFAR and remarkably better performance on ImageNet (e.g. $\mathbf{ +1.1\%}$) compared with SAM, but always require half of the training time. The code is released at https://github.com/nblt/RWP.

Randomized Smoothing (RS) is a promising technique for certified robustness, and recently in RS the ensemble of multiple deep neural networks (DNNs) has shown state-of-the-art performances. However, such an ensemble brings heavy computation burdens in both training and certification, and yet under-exploits individual DNNs and their mutual effects, as the communication between these classifiers is commonly ignored in optimization. In this work, starting from a single DNN, we augment the network with multiple heads, each of which pertains a classifier for the ensemble. A novel training strategy, namely Self-PAced Circular-TEaching (SPACTE), is proposed accordingly. SPACTE enables a circular communication flow among those augmented heads, i.e., each head teaches its neighbor with the self-paced learning using smoothed losses, which are specifically designed in relation to certified robustness. The deployed multi-head structure and the circular-teaching scheme of SPACTE jointly contribute to diversify and enhance the classifiers in augmented heads for ensemble, leading to even stronger certified robustness than ensembling multiple DNNs (effectiveness) at the cost of much less computational expenses (efficiency), verified by extensive experiments and discussions.

The LIGO, Virgo, and KAGRA (LVK) collaboration has announced 90 coalescing binary black holes (BBHs) with $p_{\rm astro} > 50\%$ to date, however, the origin of their formation channels is still an open scientific question. Given various properties of BBHs (BH component masses and individual spins) inferred using the default priors by the LVK, independent groups have been trying to explain the formation of the BBHs with different formation channels. Of all formation scenarios, the chemically homogeneous evolution (CHE) channel has stood out with distinguishing features, namely, nearly-equal component masses and preferentially high individual spins aligned with the orbital angular momentum. We perform Bayesian inference on the BBH events officially reported in GWTC-3 with astrophysically-predicted priors representing different formation channels of the isolated binary evolution (CEE: common-envelope evolution channel; CHE; SMT: stable mass transfer). Given assumed models, we report strong evidence for GW190517\_055101 being most likely to have formed through the CHE channel. Assuming the BBH events in the subsample are all formed through one of the isolated binary evolution channels, we obtain the lower limits on the local merger rate density of these channels at $11.45 ~\mathrm{Gpc^{-3}~yr^{-1}}$ (CEE), $0.18 ~\mathrm{Gpc^{-3}~yr^{-1}}$ (CHE), and $0.63 ~\mathrm{Gpc^{-3}~yr^{-1}}$ (SMT) at $90\%$ credible level., Comment: 13 pages, 4 figures, 1 table