arXiv subjects

Xiaoshan Huang

Publications and source records attributed to Xiaoshan Huang.

31 records · Page 2Linked to original sources

arXiv2025

SSRLBot: Designing and Developing a Large Language Model-based Agent using Socially Shared Regulated Learning

Large language model (LLM)--based agents have emerged as pivotal tools in assisting human experts across various fields by transforming complex tasks into more efficient workflows and providing actionable stakeholder insights. Despite their potential, the application of LLM-based agents for medical education remains underexplored. The study aims to assist in evaluating the students' process and outcomes on medical case diagnosis and discussion while incorporating the theoretical framework of Socially Shared Regulation of Learning (SSRL) to assess student performance. SSRL emphasizes metacognitive, cognitive, motivational, and emotional interactions, highlighting the collaborative management of learning processes to improve decision-making outcomes. Grounded in SSRL theory, this tool paper introduces SSRLBot, an LLM-based agent designed to enable team members to reflect on their diagnostic performance and the key SSRL skills that foster team success. SSRLBot's core functions include summarizing dialogue content, analyzing participants' SSRL skills, and evaluating students' diagnostic results. Meanwhile, we evaluated SSRLBot through diagnostic conversation data collected from six groups (12 participants, 1926 conversational turns). Results showed that SSRLBot can deliver detailed, theory-aligned evaluations, link specific behaviors to SSRL dimensions, and offer actionable recommendations for improving teamwork. The findings address a critical gap in medical education, advancing the application of LLM agents to enhance team-based decision-making and collaboration in high-stakes environments.

cs.HC↗

arXiv2025

What Makes Teamwork Work? A Multimodal Case Study on Emotions and Diagnostic Expertise in an Intelligent Tutoring System

Teamwork is pivotal in medical teamwork when professionals with diverse skills and emotional states collaborate to make critical decisions. This case study examines the interplay between emotions and professional skills in group decision-making during collaborative medical diagnosis within an Intelligent Tutoring System (ITS). By comparing verbal and physiological data between high-performing and low-performing teams of medical professionals working on a patient case within the ITS, alongside individuals' retrospective collaboration experiences, we employ multimodal data analysis to identify patterns in team emotional climate and their impact on diagnostic efficiency. Specifically, we investigate how emotion-driven dialogue and professional expertise influence both the information-seeking process and the final diagnostic decisions. Grounded in the socially shared regulation of learning framework and utilizing sentiment analysis, we found that social-motivational interactions are key drivers of a positive team emotional climate. Furthermore, through content analysis of dialogue and physiological signals to pinpoint emotional fluctuations, we identify episodes where knowledge exchange and skill acquisition are most likely to occur. Our findings offer valuable insights into optimizing group collaboration in medical contexts by harmonizing emotional dynamics with adaptive strategies for effective decision-making, ultimately enhancing diagnostic accuracy and teamwork effectiveness.

cs.HC↗

arXiv2025

Disk draining in LIGO progenitor black hole binaries and its significance to electromagnetic counterparts

The effect of tidal forces on transport within a relic accretion disk in binary black holes is studied here with a suite of two-dimensional hydrodynamic simulations. As the binary contracts due to the emission of gravitational waves, the accretion disk is truncated, and a two-armed spiral wave is excited, which remains stationary in the rotating reference frame of the coalescing binary. Such spiral waves lead to increased transport of mass and angular momentum. Our findings suggest that even in the case of weakly ionized accretion disks, spiral density waves will drain the disk long before the orbit of the two black holes decays enough for them to merge, thus dimming prospects for a detectable electromagnetic counterpart.

astro-ph.HE↗

arXiv2024

Transients by Black Hole Formation from Red Supergiants: Impact of Dense Circumstellar Matter

Failed supernovae (SNe), which are likely the main channel for forming stellar-mass black holes, are predicted to accompany mass ejections much weaker than typical core-collapse SNe. We conduct a grid of one-dimensional radiation hydrodynamical simulations to explore the emission of failed SNe from red supergiant progenitors, leveraging recent understanding of the weak explosion and the dense circumstellar matter (CSM) surrounding these stars. We find from these simulations and semi-analytical modeling that diffusion in the CSM prolongs the early emission powered by shock breakout/cooling. The early emission has peak luminosities of $\sim 10^7$-$10^8~L_\odot$ in optical and UV, and durations of days to weeks. The presence of dense CSM aids detection of the early bright peak from these events via near-future wide-field surveys such as Rubin Observatory, ULTRASAT and UVEX.

astro-ph.HE↗

arXiv2024

Automatic deductive coding in discourse analysis: an application of large language models in learning analytics

Deductive coding is a common discourse analysis method widely used by learning science and learning analytics researchers for understanding teaching and learning interactions. It often requires researchers to manually label all discourses to be analyzed according to a theoretically guided coding scheme, which is time-consuming and labor-intensive. The emergence of large language models such as GPT has opened a new avenue for automatic deductive coding to overcome the limitations of traditional deductive coding. To evaluate the usefulness of large language models in automatic deductive coding, we employed three different classification methods driven by different artificial intelligence technologies, including the traditional text classification method with text feature engineering, BERT-like pretrained language model and GPT-like pretrained large language model (LLM). We applied these methods to two different datasets and explored the potential of GPT and prompt engineering in automatic deductive coding. By analyzing and comparing the accuracy and Kappa values of these three classification methods, we found that GPT with prompt engineering outperformed the other two methods on both datasets with limited number of training samples. By providing detailed prompt structures, the reported work demonstrated how large language models can be used in the implementation of automatic deductive coding.

cs.CL↗

arXiv2024

Pre-peak Emission in Tidal Disruption Events

The rising part of a tidal disruption event light curve provides unique insight into early emission and the onset of accretion. Various mechanisms are proposed to explain the pre-peak emission, including shocks from debris interaction and reprocessing of disk emission. We study the pre-peak emission and its influence on the gas circularization by a series of gray radiation hydrodynamic simulations with varying black hole mass. We find that given a super-Eddington fallback rate of 10\dot{M}_{Edd}, the stream-stream collision can occur multiple times and drive strong outflows of up to 9\dot{M}_{Edd}. By dispersing gas to \gtrsim 100rs, the outflow can delay gas circularization and leads to sub-Eddington accretion rates during the first few stream-stream collisions. The stream-stream collision shock and circularization shock can sustain a luminosity of ~10^{44}erg/s for days. The luminosity is generally sub-Eddington and shows a weak correlation with accretion rate at early time. The outflow is optically thick, yielding a reprocessing layer with a size of ~10^{14} cm and photospheric temperature of ~4\times10^{4}K.

astro-ph.HE↗

arXiv2024

Examining the Role of Peer Acknowledgements on Social Annotations: Unraveling the Psychological Underpinnings

This study explores the impact of peer acknowledgement on learner engagement and implicit psychological attributes in written annotations on an online social reading platform. Participants included 91 undergraduates from a large North American University. Using log file data, we analyzed the relationship between learners' received peer acknowledgement and their subsequent annotation behaviours using cross-lag regression. Higher peer acknowledgements correlate with increased initiation of annotations and responses to peer annotations. By applying text mining techniques and calculating Shapley values to analyze 1,969 social annotation entries, we identified prominent psychological themes within three dimensions (i.e., affect, cognition, and motivation) that foster peer acknowledgment in digital social annotation. These themes include positive affect, openness to learning and discussion, and expression of motivation. The findings assist educators in improving online learning communities and provide guidance to technology developers in designing effective prompts, drawing from both implicit psychological cues and explicit learning behaviours.

cs.HC↗

arXiv2023

A Bright First Day for Tidal Disruption Event

Stream-stream collision may be an important pre-peak energy dissipation mechanism in tidal disruption events (TDEs). We perform local three-dimensional radiation hydrodynamic simulations in a wedge geometry including the gravity to study stream self-crossing, with emphasis on resolving the collision and following the subsequent outflow. We find that the collision can contribute to pre-peak optical emissions by converting $\gtrsim5\%$ of stream kinetic energy to radiation, yielding prompt emission of $\sim10^{42-44}\rm erg~s^{-1}$. The radiative efficiency is sensitive to stream mass fallback rates, and strongly depends on the downstream gas optical depth. Even for a sub-Eddington ($10\%$) mass fallback rate, the strong radiation pressure produced in the collision can form a local super-Eddington region near the collision site, where a fast, aspherical outflow is launched. Higher mass fallback rate usually leads to more optically-thick outflow and lower net radiative efficiency. For $\dot{M}\gtrsim0.1\dot{M}_{\rm Edd}$, the estimated photosphere size of the outflow can expand by one to two orders of magnitudes reaching $\sim10^{14}\rm cm$. The average gas temperature at this photospheric surface is a few $\times10^{4}$K, roughly consistent with inferred pre-peak photosphere properties for some optical TDEs. We find that the dynamics is sensitive to collision angle and collision radius, but the radiative efficiency or outflow properties show more complex dependency than is often assumed in ballistic models.

astro-ph.HE↗

arXiv2022

CR Driven Multi-phase Gas Formed via Thermal Instability

Cosmic rays (CRs) are an important energy source in the circum-galactic medium (CGM) that impact the multi-phase gas structure and dynamics. We perform two-dimensional CR-magnetohydrodynamic simulations to investigate the role of CRs in accelerating multi-phase gas formed via thermal instability. We compare outflows driven by CRs to those driven by a hot wind with equivalent momentum. We find that CRs driven outflow produces lower density contrast between cold and hot gas due to non-thermal pressure support, and yields a more filamentary cloud morphology. While entrainment in a hot wind can lead to cold gas increasing due to efficient cooling, CRs tend to suppress cold gas growth. The mechanism of this suppression depends on magnetic field strength, with CRs either reducing cooling or shredding the clouds by differential acceleration. Despite the suppression of cold gas growth, CRs are able to launch the cold clouds to observed velocities without rapid destruction. The dynamical interaction between CRs ad multi-phase gas is also sensitive to the magnetic field strength. In relatively strong fields, the CRs are more important for direct momentum input to cold gas. In relatively weak fields, the CRs impact gas primarily by heating, which modifies gas pressure.

astro-ph.GA↗

arXiv2022

The Launching of Cosmic Ray Driven Outflows

Cosmic rays (CRs) are thought to be an important feedback mechanism in star-forming galaxies. They can provide an important source of pressure support and possibly drive outflows. We perform multidimensional CR-magnetohydrodynamic simulations including transport by streaming and diffusion to investigate wind launching from an initially hydrostatic atmosphere by CRs. We estimate a characteristic Eddington limit on the CR flux for which the CR force exceeds gravity and compare it to simulated systems. Scaling our results to conditions in star-forming galaxies, we find that CRs are likely to contribute to driving outflows for a broad range of star formation environments. We quantify the momentum and energy transfer between CRs and gas, along with the associated mass outflow rates under different assumptions about the relative importance of streaming and diffusion for transport. In simulations with streaming, we observe the growth and saturation of the CR acoustic instability, but the CRs and gas remain well coupled, with CR momentum transferred efficiently to the gas even when this instability is present. Higher CR fluxes transferr more energy to the gas and drive stronger outflows. When streaming is present, most of the transferred energy takes the form of Alfvén wave heating of the gas, raising its pressure and internal energy, with a lower fractional contribution to the kinetic energy of the outflow. We also consider runs with radiative cooling, which modifies gas temperature and pressure profiles but does not seem to have a large impact on the mass outflow for super-Eddington CR fluxes.

astro-ph.GA↗

arXiv2021

Interpretability in deep learning for finance: a case study for the Heston model

Deep learning is a powerful tool whose applications in quantitative finance are growing every day. Yet, artificial neural networks behave as black boxes and this hinders validation and accountability processes. Being able to interpret the inner functioning and the input-output relationship of these networks has become key for the acceptance of such tools. In this paper we focus on the calibration process of a stochastic volatility model, a subject recently tackled by deep learning algorithms. We analyze the Heston model in particular, as this model's properties are well known, resulting in an ideal benchmark case. We investigate the capability of local strategies and global strategies coming from cooperative game theory to explain the trained neural networks, and we find that global strategies such as Shapley values can be effectively used in practice. Our analysis also highlights that Shapley values may help choose the network architecture, as we find that fully-connected neural networks perform better than convolutional neural networks in predicting and interpreting the Heston model prices to parameters relationship.

q-fin.PR↗

arXiv2020

Dusty Cloud Acceleration with Multiband Radiation

We perform two-dimensional and three-dimensional simulations of cold, dense clouds, which are accelerated by radiation pressure on dust relative to a hot, diffuse background gas. We examine the relative effectiveness of acceleration by ultraviolet and infrared radiation fields, both independently and acting simultaneously on the same cloud. We study clouds that are optically thin to infrared emission but with varying ultraviolet optical depths. Consistent with previous work, we find relatively efficient acceleration and long cloud survival times when the infrared band flux dominates over the ultraviolet flux. However, when ultraviolet is dominant or even a modest percentage ($\sim 5-10$\%) of the infrared irradiating flux, it can act to compress the cloud, first crushing it and then disrupting the outer layers. This drives mixing of outer regions of the dusty gas with the hot diffuse background to the point where most dust is not likely to survive or stay coupled to the gas. Hence, the cold cloud is unable to survive for a long enough timescale to experience significant acceleration before disruption even though efficient infrared cooling keeps the majority of the gas close to radiative equilibrium temperature ($T \lesssim 100$K). We discuss implications for observed systems, concluding that radiation pressure driving is most effective when the light from star-forming regions is efficiently reprocessed into the infrared.

astro-ph.GA↗

arXiv2020

BETA: A Large Benchmark Database Toward SSVEP-BCI Application

Brain-computer interface (BCI) provides an alternative means to communicate and it has sparked growing interest in the past two decades. Specifically, for Steady-State Visual Evoked Potential based BCI, marked improvement has been made in the frequency recognition method and data sharing. However, the number of pubic database is still limited in this field. Therefore, we present a \textbf{BE}nchmark database \textbf{T}owards BCI \textbf{A}pplication (BETA) in the study. The BETA database is composed of 64-channel Electroencephalogram (EEG) data from 70 subjects performing a 40-target cued-spelling task. The design and acquisition of BETA is in pursuit of meeting the demand from real-world applications and it can be used as a test-bed for these scenarios. We validate the database by a series of analysis and conduct the classification analysis of eleven frequency recognition methods on BETA. We recommend to use the metric of wide-band SNR and BCI quotient to characterize the SSVEP at the single-trial and population level, respectively. The BETA database can be downloaded from the website http://bci.med.tsinghua.edu.cn/download.html.

eess.SP↗