arXiv subjects

S. Bai

Publications and source records attributed to S. Bai.

8 recordsLinked to original sources

arXiv2026

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↗

arXiv2026

MiroThinker-1.7 & H1: Towards Heavy-Duty Research Agents via Verification

We present MiroThinker-1.7, a new research agent designed for complex long-horizon reasoning tasks. Building on this foundation, we further introduce MiroThinker-H1, which extends the agent with heavy-duty reasoning capabilities for more reliable multi-step problem solving. In particular, MiroThinker-1.7 improves the reliability of each interaction step through an agentic mid-training stage that emphasizes structured planning, contextual reasoning, and tool interaction. This enables more effective multi-step interaction and sustained reasoning across complex tasks. MiroThinker-H1 further incorporates verification directly into the reasoning process at both local and global levels. Intermediate reasoning decisions can be evaluated and refined during inference, while the overall reasoning trajectory is audited to ensure that final answers are supported by coherent chains of evidence. Across benchmarks covering open-web research, scientific reasoning, and financial analysis, MiroThinker-H1 achieves state-of-the-art performance on deep research tasks while maintaining strong results on specialized domains. We also release MiroThinker-1.7 and MiroThinker-1.7-mini as open-source models, providing competitive research-agent capabilities with significantly improved efficiency.

cs.CL↗

arXiv2024

Existence and multiplicity of solutions for critical Kirchhoff-Choquard equations involving the fractional $p$-Laplacian on the Heisenberg group

In this paper, we study existence and multiplicity of solutions for the following Kirchhoff-Choquard type equation involving the fractional $p$-Laplacian on the Heisenberg group: \begin{equation*} \begin{array}{lll} M(\|u\|_{\mu}^{p})(\mu(-\Delta)^{s}_{p}u+V(\xi)|u|^{p-2}u)= f(\xi,u)+\int_{\mathbb{H}^N}\frac{|u(\eta)|^{Q_{\lambda}^{\ast}}}{|\eta^{-1}\xi|^{\lambda}}d\eta|u|^{Q_{\lambda}^{\ast}-2}u &\mbox{in}\ \mathbb{H}^N, \\ \end{array} \end{equation*} where $(-\Delta)^{s}_{p}$ is the fractional $p$-Laplacian on the Heisenberg group $\mathbb{H}^N$, $M$ is the Kirchhoff function, $V(\xi)$ is the potential function, $0 0$, $f(\xi,u)$ is the nonlinear function, $0<\lambda<Q$, $Q=2N+2$, and $Q_{\lambda}^{\ast}=\frac{2Q-\lambda}{Q-2}$ is the Sobolev critical exponent. Using the Krasnoselskii genus theorem, the existence of infinitely many solutions is obtained if $\mu$ is sufficiently large. In addition, using the fractional version of the concentrated compactness principle, we prove that problem has $m$ pairs of solutions if $\mu$ is sufficiently small. As far as we know, the results of our study are new even in the Euclidean case.

math.AP↗

arXiv2020

An agent-based negotiation model and its implementation in Repast

We propose an agent-based model, MNegoti, for simulating multilateral negotiation process, which can be naturally employed in group decision support system. This model can also be applied to any use case in which negotiation is involved, in order to simulate the negotiation process. In this report, we discuss the implementation of the MNegoti model on the basis of the agent-based simulation platform, Repast Simphony. It is worth pointing out that this model can be used to create a java module for any use of agent-based negotiation simulation.

cs.MA↗

arXiv2020

Gamma Ray Flashes Produced by Lightning Observed at Ground Level by TETRA-II

In its first 2 years of operation, the ground-based Terrestrial gamma ray flash and Energetic Thunderstorm Rooftop Array(TETRA)-II array of gamma ray detectors has recorded 22 bursts of gamma rays of millisecond-scale duration associated with lightning. In this study, we present the TETRA-II observations detected at the three TETRA-II ground-level sites in Louisiana, Puerto Rico, and Panama together with the simultaneous radio frequency signals from the VAISALA Global Lightning Data set, VAISALA National Lightning Detection Network, Earth Networks Total Lightning Network, and World Wide Lightning Location Network. The relative timing between the gamma ray events and the lightning activity is a key parameter for understanding the production mechanism(s) of the bursts. The gamma ray time profiles and their correlation with radio sferics suggest that the gamma ray events are initiated by lightning leader activity and are produced near the last stage of lightning leader channel development prior to the lightning return stroke.

physics.ao-ph↗

arXiv2018

Evaluation of Beam Halo from Beam-Gas Scattering at the KEK-ATF

In circular colliders, as well as in damping rings and synchrotron radiation light sources, beam halo is one of the critical issues limiting the performance as well as potentially causing component damage and activation. It is imperative to clearly understand the mechanisms that lead to halo formation and to test the available theoretical models. Elastic beam-gas scattering can drive particles to large oscillation amplitudes and be a potential source of beam halo. In this paper, numerical estimation and Monte Carlo simulations of this process at the ATF of KEK are presented. Experimental measurements of beam halo in the ATF2 beam line using a diamond sensor detector are also described, which clearly demonstrates the influence of the beam-gas scattering process on the transverse halo distribution.

physics.acc-ph↗

arXiv2016

Numerical study of the Kerr solution in rotating coordinates

The Kerr solution in coordinates corotating with the horizon is studied as a testbed for a spacetime with a helical Killing vector in the Ernst picture. The solution is numerically constructed by solving the Ernst equation with a spectral method and a Newton iteration. We discuss convergence of the iteration for several initial iterates and different values of the Kerr parameters.

gr-qc↗

arXiv2012

Present status and first results of the final focus beam line at the KEK Accelerator Test Facility

ATF2 is a final-focus test beam line which aims to focus the low emittance beam from the ATF damping ring to a vertical size of about 37 nm and to demonstrate nanometer level beam stability. Several advanced beam diagnostics and feedback tools are used. In December 2008, construction and installation were completed and beam commissioning started, supported by an international team of Asian, European, and U.S. scientists. The present status and first results are described.

physics.acc-ph↗