arXiv subjects

Woonghee Han

Publications and source records attributed to Woonghee Han.

3 recordsLinked to original sources

arXiv2025

Modeling User-System Behavior for Training-free Building of Private Domain Conversational Agents

The rise of agentic systems that combine orchestration, tool use, and conversational capabilities, has been more visible by the recent advent of large language models (LLMs). While open-domain frameworks exist, applying them in private domains remains difficult due to heterogeneous tool formats, domain-specific jargon, restricted accessibility of APIs, and complex governance. Conventional solutions, such as fine-tuning on synthetic dialogue data, are burdensome and brittle under domain shifts, and risk degrading general performance. In this light, we introduce a framework for constructing private-domain tool-calling multi-agent systems that does not necessitate data generation and model tuning, by modeling user and system behaviors via tool-use documentation. With a simplified composition that assumes an orchestrator, a tool-calling agent, and a general chat agent, we show that the proposed approach enables scalable adaptation to private tools, which had been previously achieved by accompanying continual retraining. Our framework supports practical use cases, including lightweight deployment of in-house multi-agent systems, providing a sustainable method for aligning agent behavior with domain expertise in private, enterprise-level conversational ecosystems.

cs.MA↗

arXiv2022

X-point and divertor filament dynamics from Gas Puff Imaging on TCV

A new Gas Puff Imaging (GPI) diagnostic has been installed on the TCV tokamak, providing two-dimensional insights into Scrape-Off-Layer (SOL) turbulence dynamics above, at and below the magnetic X-point. A detailed study in L-mode, attached, lower single-null discharges shows that statistical properties have little poloidal variations, while vast differences are present in the 2D behaviour of intermittent filaments. Strongly elongated filaments, just above the X-point and in the divertor far-SOL, show a good consistency in shape and dynamics with field-line tracing from filaments at the outboard midplane, highlighting their connection. In the near-SOL of the outer divertor leg, shortlived, high frequency and more circular (diameter $\sim$15 sound Larmour radii) filaments are observed. These divertor-localised filaments appear born radially at the position of maximum density and display a radially outward motion with velocity $\approx$400\,m/s that is comparable to radial velocities of upstream-connected filaments. Conversely, in these discharges ($B\times\nabla B$ pointing away from the divertor), these divertor filaments' poloidal velocities differ strongly from those of upstream-connected filaments. The importance of divertor-localised filaments upon radial transport and profile broadening is explored using filament statistics and in-situ kinetic profile measurements along the divertor leg. This provides evidence that these filaments contribute significantly to electron density profile broadening in the divertor.

physics.plasm-ph↗

arXiv2021

Tracking Blobs in the Turbulent Edge Plasma of a Tokamak Fusion Device

The analysis of turbulence in plasmas is fundamental in fusion research. Despite extensive progress in theoretical modeling in the past 15 years, we still lack a complete and consistent understanding of turbulence in magnetic confinement devices, such as tokamaks. Experimental studies are challenging due to the diverse processes that drive the high-speed dynamics of turbulent phenomena. This work presents a novel application of motion tracking to identify and track turbulent filaments in fusion plasmas, called blobs, in a high-frequency video obtained from Gas Puff Imaging diagnostics. We compare four baseline methods (RAFT, Mask R-CNN, GMA, and Flow Walk) trained on synthetic data and then test on synthetic and real-world data obtained from plasmas in the Tokamak `a Configuration Variable (TCV). The blob regime identified from an analysis of blob trajectories agrees with state-of-the-art conditional averaging methods for each of the baseline methods employed, giving confidence in the accuracy of these techniques. High entry barriers traditionally limit tokamak plasma research to a small community of researchers in the field. By making a dataset and benchmark publicly available, we hope to open the field to a broad community in science and engineering.

physics.plasm-ph↗