Search

3D deblurring reconstruction techniques have recently seen significant advancements with the development of Neural Radiance Fields (NeRF) and 3D Gaussian Splatting (3DGS). Although these techniques can recover relatively clear 3D reconstructions from blurry image inputs, they still face limitations in handling severe blurring and complex camera motion. To address these issues, we propose Event-assisted 3D Deblur Reconstruction with Gaussian Splatting (EaDeblur-GS), which integrates event camera data to enhance the robustness of 3DGS against motion blur. By employing an Adaptive Deviation Estimator (ADE) network to estimate Gaussian center deviations and using novel loss functions, EaDeblur-GS achieves sharp 3D reconstructions in real-time, demonstrating performance comparable to state-of-the-art methods.

Quantum secure direct communication (QSDC) is a approach of communication to transmit secret messages base on quantum mechanics. Different from the quantum key distribution, secret messages can be transmitted directly on quantum channel with QSDC. Higher channel capacity and noise suppression capabilities are key to achieving long-distance quantum communication. Here we report a continuous-variable QSDC scheme based on mask-coding and orbital angular momentum, in which the mask-coding is employed to protect the security of the transmitting messages and to suppression the influence of excess noise. And the combination of orbital angular momentum and the information block transmission can effectively improve the secrecy capacity. In the 800 information blocks * 1310 bits length 10-km experiment, the results show that the statistical average of bit error rate has decreased by 36% and reached 0.38% comparing to existing solutions, and the system excess noise is 0.0184 SNU, and the final secrecy capacity is 6.319 * 106 bps. Therefore, this scheme reduces error bits while increasing secrecy capacity, providing a solution for long-distance large-scale quantum communication.

Leveraging scattering information to describe binary systems in generic orbits requires identifying local- and nonlocal-in-time tail effects. We report here the derivation of the universal (non-spinning) local-in-time conservative dynamics at fourth Post-Minkowskian order, i.e. ${\cal O}(G^4)$. This is achieved by computing the nonlocal-in-time contribution to the deflection angle, and removing it from the full conservative value in [2112.11296,2210.05541]. Unlike the total result, the integration problem involves two scales, velocity and mass ratio, and features multiple polylogarithms, complete elliptic and iterated elliptic integrals, notably in the mass ratio. We reconstruct the local radial action, center-of-mass momentum and Hamiltonian, as well as the exact logarithmic-dependent part(s), all valid for generic orbits. We incorporate the remaining nonlocal terms for elliptic-like motion to sixth Post-Newtonian order. The combined Hamiltonian is in perfect agreement in the overlap with the Post-Newtonian state of the art. The results presented here provide the most accurate description of gravitationally-bound binaries harnessing scattering data to date, readily applicable to waveform modelling., Comment: 5 pages + Refs + Supplemental. Two computer-readable ancillary files. Two computer-readable ancillary files. v2: Ancillary files corrected to be compatible with the split in Eq. (2) and the result quoted in Eq. (7) (unmodified) with the given h_i coefficients (unmodified). To appear in Phys. Rev. Lett

We investigate the amplification of the genuine tripartite nonlocality(GTN) and the genuine tripartite entanglement(GTE) of Dirac particles in the background of a Schwarzschild black hole by a local filtering operation under decoherence. It is shown that the physically accessible GTN will be completely destroyed by decoherence, which means that the physically accessible GTN will not exist in the system. Particularly, the local filtering operation can make the physically accessible GTN appear within a certain range of Hawking temperature, namely, the local filtering operation can cause the physically accessible GTN to be generated in the system coupled with the environment, which is not discovered before and is benefit for the quantum information processing. Furthermore, we also find that the physically accessible GTE approaches a stable value in the limit of infinite Hawking temperature for most cases, but if the decoherence parameter $p$ is less than 1, the ``sudden death'' of GTE will take place when the decoherence strength is large enough. It is worth noting that the nonzero stable value of GTE can be increased by performing the local filtering operation, even in the presence of decoherence. Finally, we explore the generation of physically inaccessible GTN and GTE of other tripartite subsystems under decoherence, it is shown that the physically inaccessible GTN cannot be produced, but the physically inaccessible GTE can be produced. In addition, we can see that the generated physically inaccessible GTE can be increased by applying the local filtering operation., Comment: 11 pages, 9 figures. arXiv admin note: text overlap with arXiv:2201.02333 by other authors

Intelligent agents stand out as a potential path toward artificial general intelligence (AGI). Thus, researchers have dedicated significant effort to diverse implementations for them. Benefiting from recent progress in large language models (LLMs), LLM-based agents that use universal natural language as an interface exhibit robust generalization capabilities across various applications -- from serving as autonomous general-purpose task assistants to applications in coding, social, and economic domains, LLM-based agents offer extensive exploration opportunities. This paper surveys current research to provide an in-depth overview of LLM-based intelligent agents within single-agent and multi-agent systems. It covers their definitions, research frameworks, and foundational components such as their composition, cognitive and planning methods, tool utilization, and responses to environmental feedback. We also delve into the mechanisms of deploying LLM-based agents in multi-agent systems, including multi-role collaboration, message passing, and strategies to alleviate communication issues between agents. The discussions also shed light on popular datasets and application scenarios. We conclude by envisioning prospects for LLM-based agents, considering the evolving landscape of AI and natural language processing.

Modern displays can render video content with high dynamic range (HDR) and wide color gamut (WCG). However, most resources are still in standard dynamic range (SDR). Therefore, transforming existing SDR content into the HDRTV standard holds significant value. This paper defines and analyzes the SDRTV-to-HDRTV task by modeling the formation of SDRTV/HDRTV content. Our findings reveal that a naive endto-end supervised training approach suffers from severe gamut transition errors. To address this, we propose a new three-step solution called HDRTVNet++, which includes adaptive global color mapping, local enhancement, and highlight refinement. The adaptive global color mapping step utilizes global statistics for image-adaptive color adjustments. A local enhancement network further enhances details, and the two sub-networks are combined as a generator to achieve highlight consistency through GANbased joint training. Designed for ultra-high-definition TV content, our method is both effective and lightweight for processing 4K resolution images. We also constructed a dataset using HDR videos in the HDR10 standard, named HDRTV1K, containing 1235 training and 117 testing images, all in 4K resolution. Additionally, we employ five metrics to evaluate SDRTV-to-HDRTV performance. Our results demonstrate state-of-the-art performance both quantitatively and visually. The codes and models are available at https://github.com/xiaom233/HDRTVNet-plus., Comment: Extended version of HDRTVNet

The dual-carbon goals of China necessitate precise accounting of company carbon emissions, vital for green development across all industries. Not only the company itself but also financial investors require accurate and comprehensive company-level emissions data for climate risk management. This paper introduces the structure and methodology of the High-resolution Database for Carbon Emissions of China-listed companies, integrating three primary data sources: self-disclosed environmental data from listed companies, long-accumulated national power emission data, and regional high-precision emission data derived from multi-source satellites. The database's innovation lies in the employment of artificial intelligence (AI) algorithms to aggregate multi-source satellite data. This approach enables the precise identification of carbon emission sources and the prediction of company-level carbon emissions. Consequently, this methodology robustly cross-validates self-reported direct emissions, enhancing the accuracy and granularity of company-level emission records. Central to the database's utility includes the provision of high-resolution company carbon emission data, which is not only highly accurate but also instrumental in carbon management and emission market transactions. By offering a more nuanced and verifiable picture of company emissions, the database supports China's broader efforts to meet its ambitious dual-carbon targets and transition towards a more sustainable and environmentally responsible economy.

Colorectal cancer is one of the most common cancers in the world. While colonoscopy is an effective screening technique, navigating an endoscope through the colon to detect polyps is challenging. A 3D map of the observed surfaces could enhance the identification of unscreened colon tissue and serve as a training platform. However, reconstructing the colon from video footage remains difficult. Learning-based approaches hold promise as robust alternatives, but necessitate extensive datasets. Establishing a benchmark dataset, the 2022 EndoVis sub-challenge SimCol3D aimed to facilitate data-driven depth and pose prediction during colonoscopy. The challenge was hosted as part of MICCAI 2022 in Singapore. Six teams from around the world and representatives from academia and industry participated in the three sub-challenges: synthetic depth prediction, synthetic pose prediction, and real pose prediction. This paper describes the challenge, the submitted methods, and their results. We show that depth prediction from synthetic colonoscopy images is robustly solvable, while pose estimation remains an open research question.

We describe the formalism to compute gravitational-wave observables for compact binaries via the effective field theory framework in combination with modern tools from collider physics. We put particular emphasis on solving the "multi-loop" integration problem via the methodology of differential equations and expansion by regions. This allows us to "bootstrap" the two-body relativistic dynamics in the Post-Minkowskian (PM) expansion from boundary data evaluated in the near-static (soft) limit. We illustrate the procedure with the derivation of the total spacetime impulse in the scattering of non-spinning bodies to 4PM (three-loop) order, i.e. ${\cal O}(G^4)$, including conservative and dissipative effects., Comment: 50 pages + Apps and Refs (89 total). 1 computer-readable ancillary file

To limit global warming to pre-industrial levels, global governments, industry and academia are taking aggressive efforts to reduce carbon emissions. The evaluation of anthropogenic carbon dioxide (CO$_2$) emissions, however, depends on the self-reporting information that is not always reliable. Society need to develop an objective, independent, and generalized system to meter CO$_2$ emissions. Satellite CO$_2$ observation from space that reports column-average regional CO$_2$ dry-air mole fractions has gradually indicated its potential to build such a system. Nevertheless, estimating anthropogenic CO$_2$ emissions from CO$_2$ observing satellite is bottlenecked by the influence of the highly complicated physical characteristics of atmospheric activities. Here we provide the first method that combines the advanced artificial intelligence (AI) techniques and the carbon satellite monitor to quantify anthropogenic CO$_2$ emissions. We propose an integral AI based pipeline that contains both a data retrieval algorithm and a two-step data-driven solution. First, the data retrieval algorithm can generate effective datasets from multi-modal data including carbon satellite, the information of carbon sources, and several environmental factors. Second, the two-step data-driven solution that applies the powerful representation of deep learning techniques to learn to quantify anthropogenic CO$_2$ emissions from satellite CO$_2$ observation with other factors. Our work unmasks the potential of quantifying CO$_2$ emissions based on the combination of deep learning algorithms and the carbon satellite monitor.

We obtain the total impulse in the scattering of non-spinning binaries in general relativity at fourth Post-Minkowskian order, i.e. ${\cal O}(G^4)$, including linear, nonlinear, and hereditary radiation-reaction effects. We derive the total radiated spacetime momentum as well as the associated energy flux. The latter can be used to compute gravitational-wave observables for generic (un)bound orbits. We employ the ("in-in") Schwinger-Keldysh worldline effective field theory framework in combination with modern "multi-loop" integration techniques from collider physics. The complete results are in agreement with various partial calculations in the Post-Newtonian/Minkowskian expansion., Comment: 6 pages + Refs + Supplemental. 1 figure and 1 table. 1 computer-readable ancillary file

We study a neural network framework for the numerical evaluation of Feynman loop integrals that are fundamental building blocks for perturbative computations of physical observables in gauge and gravity theories. We show that such a machine learning approach improves the convergence of the Monte Carlo algorithm for high-precision evaluation of multi-dimensional integrals compared to traditional algorithms. In particular, we use a neural network to improve the importance sampling. For a set of representative integrals appearing in the computation of the conservative dynamics for a compact binary system in General Relativity, we perform a quantitative comparison between the Monte Carlo integrators VEGAS and i-flow, an integrator based on neural network sampling., Comment: 26 pages + references, 4 figures, 3 tables, added ancillary, journal version