SearcharxivSearch

arXiv subjects

S. Du

Publications and source records attributed to S. Du.

7 recordsLinked to original sources

Apodex 1.1: Scaling Agentic Intelligence for Complex Work

General-purpose language models can reason and synthesize knowledge, but complex work also requires sustained interaction with files, information sources, and executable code, together with state maintenance, failure recovery, and verifiable delivery. We call this \emph{working capability}: sustained, verifiable progress toward a real-world objective. Apodex 1.1 develops this capability along two complementary dimensions. \emph{Environment Scaling} expands the diversity and verifiability of executable file, search, and code environments, while \emph{Agentic Coordination Scaling} trains agents to decompose long-horizon tasks, delegate parallel work, integrate asynchronous results, and replan. A shared execution harness and AgentOS maintain task state and provenance across tools and agents, and training turns environment trajectories and coordination traces into reliable behavior. Across complex professional work, finance, scientific research, mathematics, coding, and search, Apodex 1.1 reaches the leading performance band despite using a substantially smaller model than many frontier systems. The 35B-parameter Apodex 1.1 Mini further retains strong working capability in a locally deployable form. These results ground agentic intelligence in useful, verifiable work completed over time and advance our goal of building a \emph{Heavy-Duty Solver} for ambitious, long-running tasks.

cs.AI

The SVOM French Science Center Infrastructure

At the heart of the SVOM French ground segment, the French Science Center is a cloud-based platform which provides services and tools for the management, storage, scientific processing and visualization of SVOM data for the French community. This digital center is a critical node of the SVOM system since it is the single point of access to SVOM scientific data for the French community and the only component of the ground system connected to the SVOM VHF network allowing near-time communication from the satellite. Scientific processing pipelines are fully integrated into its infrastructure, allowing the automated production of high-level scientific data and the generation and broadcast of alerts to the scientific community at large. The software components of the French Science Center are running 24/7 and thus require a high level of automation, which led to the development of dedicated software relying on modern technical solutions such as micro-services, application containerization, infrastructure-as-code and continuous integration and deployment. In this paper, we describe the FSC infrastructure design, technological choices, and the process of SVOM data ingestion, archiving, and automated processing by the FSC scientific pipelines. We present analysis on the FSC performances in terms of availability, amount of data processed as well as processing speed.

astro-ph.IM

Dynamics of the Upwind Heliosphere Due to Data-Driven, Solar Wind and Magnetic Field Variations and Implications for Wave Propagation into the Very Local Interstellar Medium

We introduce an updated, time-dependent treatment to the split-tail ("croissant-like") heliosphere model with data-driven solar wind conditions at 1 au, to study the evolution of the heliosphere with solar-cycle variations in plasma speed, plasma density, and magnetic field intensity. The model produces a sub-Alfv\'enic and low beta region, not observed by the Voyagers, ~15 au ahead of the heliopause. The simulated magnetic field and radial flow depart from Voyager observations in this region, indicating that time-dependent effects alone are not sufficient to understand this regime of the heliosheath. We decompose fast and slow magnetosonic wave modes from time-dependent plasma pulse structures in the heliosheath, using a linear Riemann variable analysis, for the first time. Fast mode waves can both reflect at the heliopause and transmit into the interstellar medium, and their speeds are unaffected by the low beta plasma regime in front of the heliopause. The model reproduces the pf2 jump in the magnetic field at ~2020 in the interstellar medium and we attribute the source of pressure fronts observed by Voyager 1 in the interstellar medium, and pressure pulses observed by Voyager 2 in the heliosheath, to fast mode waves that are commonly recurring solar-cycle features. The presence of fast mode waves drive a highly variable termination shock, with average radial speeds of 6.05 au yr$^{-1}$ $\pm$ 5.37 au yr$^{-1}$ in the New Horizons direction. We find that the termination shock has a sinusoidal-like oscillatory motion in the rising phase of the solar cycle, and broad inward motions during the declining phase.

astro-ph.SR

Topology Design for GNSSs Considering Both Inter-satellite Links and Ground-satellite Links

Inter-satellite links (ISLs) are adopted in global navigation satellite systems (GNSSs) for high-precision orbit determination and space-based end-to-end telemetry telecommand control and communications. Due to limited onboard ISL terminals, the polling time division duplex (PTDD) mechanism is usually proposed for space link layer networking. By extending the polling mechanism to ground-satellite links (GSLs), a unified management system of the space segment and the ground segment can be realized. However, under the polling system how to jointly design the topology of ISLs and GSLs during every slot to improve data interaction has not been studied. In this paper, we formulate the topology design problem as an integer linear programming, aiming at minimizing the average delay of data delivery from satellites to ground stations while satisfying the ranging requirement for the orbit determination. To tackle the computational complexity problem, we first present a novel modeling method of delay to reduce the number of decision variables. Further, we propose a more efficient heuristic based on maximum weight matching algorithms. Simulation results demonstrate the feasibility of the proposed methods for practical operation in GNSSs. Comparing the two methods, the heuristic can achieve similar performance with respect to average delay but with significantly less complexity.

eess.SY

Efficient quantum memory for single photon polarization qubits

A quantum memory, for storing and retrieving flying photonic quantum states, is a key interface for realizing long-distance quantum communication and large-scale quantum computation. While many experimental schemes of high storage-retrieval efficiency have been performed with weak coherent light pulses, all quantum memories for true single photons achieved so far have efficiencies far below 50%, a threshold value for practical applications. Here, we report the demonstration of a quantum memory for single-photon polarization qubits with an efficiency of >85% and a fidelity of >99 %, basing on balanced two-channel electromagnetically induced transparency in laser-cooled rubidium atoms. For the single-channel quantum memory, the optimized efficiency for storing and retrieving single-photon temporal waveforms can be as high as 90.6 %. Our result pushes the photonic quantum memory closer to its practical applications in quantum information processing.

quant-ph

Final results for the neutron $\beta$-asymmetry parameter $A_0$ from the UCNA experiment

The UCNA experiment was designed to measure the neutron $\beta$-asymmetry parameter $A_0$ using polarized ultracold neutrons (UCN). UCN produced via downscattering in solid deuterium were polarized via transport through a 7 T magnetic field, and then directed to a 1 T solenoidal electron spectrometer, where the decay electrons were detected in electron detector packages located on the two ends of the spectrometer. A value for $A_0$ was then extracted from the asymmetry in the numbers of counts in the two detector packages. We summarize all of the results from the UCNA experiment, obtained during run periods in 2007, 2008--2009, 2010, and 2011--2013, which ultimately culminated in a 0.67\% precision result for $A_0$.

nucl-ex

Tin+1Cn MXene with fully saturated and thermally stable Cl terminations

MXenes are a rapidly growing family of 2D materials that exhibit a highly versatile structure and composition, allowing for significant tuning of the material properties. These properties are, however, ultimately limited by the surface terminations, which are typically a mixture of species, including F and O that are inherent to the MXene processing. Other and robust terminations are lacking. Here, we apply high-resolution scanning transmission electron microscopy (STEM), corresponding image simulations and first-principles calculations to investigate the surface terminations on MXenes synthesized from MAX phases through Lewis acidic melts. The results show that atomic Cl terminates the synthesized MXenes, with mere residual presence of other termination species. Furthermore, in situ STEM-electron energy loss spectroscopy (EELS) heating experiments show that the Cl terminations are stable up to 750 {\deg}C. Thus, we present an attractive new termination that widely expands the MXenes functionalization space and enable new applications.

cond-mat.mtrl-sci