BZPEERReading, trust and collaboration for research
Menu

Discover

GraphFollow connections around a paper, topic or saved view.

Workspaces

HomePick up where your research flow left off.InboxHandle requests, replies and notifications from one place.CollectionsCurate reading lists, evidence sets and shareable dossiers.ResearchSee your private provenance graph, trail and imports.CollaborationMove from discovery into focused discussion rooms.PublishDraft, preview and publish full-text research articles.

Network

ExpertsFind specialists worth following or asking for help.CommunitiesJoin research circles organized around shared work.PartnersSee external organizations active around the same topics.

Opportunities

ListingsBrowse roles, calls and collaborations with clearer fit signals.

Account

Log inReturn to your reading and collaboration workspace.Create accountStart saving work, following people and joining teams.
Log inCreate account

Researcher profile

Ge Chen

Ge Chen contributes to research discovery and scholarly infrastructure.

ResearcherAffiliation not importedOpen to collaborate
math.OCastro-ph.HEeess.SYSystems and ControlArtificial Intelligenceastro-ph.GAComputation and LanguageComputer VisionInformation RetrievalMultiagent Systemsnlin.AOphysics.ao-phSocial and Information Networks

Trust snapshot

Quick read

Trust 21 - EmergingVerification L1Unclaimed author
13works
0followers
13topics
4close collaborators

Actions

Decide how to stay connected

Follow researcher0

Identity and collaboration

How to connect with this researcher

Claiming links this public author record to a researcher profile and unlocks direct collaboration workflows.

Log in to claim

Direct collaboration

Open a focused conversation when the fit is right

Claim this author entity first to unlock direct invitations.

Research graph

See the researcher in context

Open full explorer

Inspect adjacent work, topics, institutions and collaborators without jumping out to a separate graph page.

Building this graph slice

BZPEER is loading the nearby papers, people, topics and institutions for this page.

Published work

13 published item(s)

preprint2026arXiv

AMAP Agentic Planning Technical Report

We present STAgent, an agentic large language model tailored for spatio-temporal understanding, designed to solve complex tasks such as constrained point-of-interest discovery and itinerary planning. STAgent is a specialized model capable of interacting with ten distinct tools within spatio-temporal scenarios, enabling it to explore, verify, and refine intermediate steps during complex reasoning. Notably, STAgent effectively preserves its general capabilities. We empower STAgent with these capabilities through three key contributions: (1) a stable tool environment that supports over ten domain-specific tools, enabling asynchronous rollout and training; (2) a hierarchical data curation framework that identifies high-quality data like a needle in a haystack, curating high-quality queries by retaining less than 1\% of the raw data, emphasizing both diversity and difficulty; and (3) a cascaded training recipe that starts with a seed SFT stage acting as a guardian to measure query difficulty, followed by a second SFT stage fine-tuned on queries with high certainty, and an ultimate RL stage that leverages data of low certainty. Initialized with Qwen3-30B-A3B to establish a strong SFT foundation and leverage insights into sample difficulty, STAgent yields promising performance on TravelBench while maintaining its general capabilities across a wide range of general benchmarks, thereby demonstrating the effectiveness of our proposed agentic model.

0 citations0 reviewsNo code linkedOpen access
preprint2026arXiv

RAG-Enhanced Large Language Models for Dynamic Content Expiration Prediction in Web Search

In commercial web search, aligning content freshness with user intent remains challenging due to the highly varied lifespans of information. Traditional industrial approaches rely on static time-window filtering, resulting in "one-size-fits-all" rankings where content may be chronologically recent but semantically expired. To address the limitation, we present a novel Large Language Models (LLMs)-based Query-Aware Dynamic Content Expiration Prediction Framework deployed in Baidu search, reformulating timeliness as a dynamic validity inference task. Our framework extracts fine-grained temporal contexts from documents and leverages LLMs to deduce a query-specific "validity horizon"-a semantic boundary defining when information becomes obsolete based on user intent. Integrated with robust hallucination mitigation strategies to ensure reliability, our approach has been evaluated through offline and online A/B testing on live production traffic. Results demonstrate significant improvements in search freshness and user experience metrics, validating the effectiveness of LLM-driven reasoning for solving semantic expiration at an industrial scale.

0 citations0 reviewsNo code linkedOpen access
preprint2024arXiv

Multi-agent Modeling and Optimal Pumping Control of Magnetic Artificial Cilia

Tiny cilia drive the flow of surrounding fluids through asymmetric jumping, which is one of the main ways for biological organisms to control fluid transport at the micro-scale. Due to its huge application prospects in medical and environmental treatment fields, artificial cilia have attracted widespread research interest in recent years. However, how to model and optimize artificial cilia is currently a common challenge faced by scholars. We model a single artificial cilium driven by a magnetic field as a multi-agent system, where each agent is a magnetic bead, and the interactions between beads are influenced by the magnetic field. Our system is driven by controlling the magnetic field input to achieve fluid transport at low Reynolds number. In order to quantify the flow conveying capacity, we introduce the pumping performance and propose an optimal control problem for pumping performance, and then give its numerical solution. The calculation results indicate that our model and optimal control algorithm can significantly improve the pumping performance of a single cilia.

0 citations0 reviewsNo code linkedOpen access
preprint2023arXiv

Deep Synoptic Array science: a 50 Mpc fast radio burst constrains the mass of the Milky Way circumgalactic medium

We present the Deep Synoptic Array (DSA-110) discovery and interferometric localization of the so far non-repeating FRB 20220319D. The FRB originates in a young, rapidly star-forming barred spiral galaxy, IRAS 02044$+$7048, at just 50 Mpc. Although the NE2001 and YMW16 models for the Galactic interstellar-medium (ISM) contribution to the DM of FRB 20220319D exceed its total observed DM, we show that uncertainties in these models accommodate an extragalactic origin for the burst. We derive a conservative upper limit on the DM contributed by the circumgalactic medium (CGM) of the Milky Way: the limit is either 28.7 pc cm$^{-3}$ and 47.3 pc cm$^{-3}$, depending on which of two pulsars nearby on the sky to FRB 20220319D is used to estimate the ISM DM. These limits both imply that the total Galactic CGM mass is $<10^{11}M_{\odot}$, and that the baryonic mass of the Milky Way is $\lesssim60\%$ of the cosmological average given the total halo mass. More stringent albeit less conservative constraints are possible when the DMs of pulsars in the distant globular cluster M53 are additionally considered. Although our constraints are sensitive to possible anisotropy in the CGM and to the assumed form of the radial-density profile, they are not subject to uncertainties in the chemical and thermal properties of the CGM. Our results strongly support scenarios commonly predicted by galaxy-formation simulations wherein feedback processes expel baryonic matter from the halos of galaxies like the Milky Way.

0 citations0 reviewsNo code linkedOpen access
preprint2022arXiv

A comprehensive observational study of the FRB 121102 persistent radio source

FRB 121102 is the first fast radio burst source to be spatially associated with a persistent radio source (QRS121102), the nature of which remains unknown. We present a detailed observational study of QRS121102 and its host galaxy. We constrain the physical size of QRS121102 by measuring its flux-density variability with the VLA in the Ku-band (12 to 18 GHz) and the K-band (18 to 26 GHz). Any such variability would likely be due to Galactic refractive scintillation and would require the source radius to be <10^17 cm at the host-galaxy redshift. We found the radio variability to be lower than the scintillation theory predictions for such a small source, leaving open the possibility for non-AGN models for QRS121102. In addition, we roughly estimated the mass of any potential supermassive black hole (SMBH) associated with QRS121102 from the width of the Hαemission line using a medium-resolution optical spectrum from the Keck Observatory. The line width gives a velocity dispersion of <30 km/s, indicating a SMBH mass of <10^{4~5} M_sun. We find the SMBH mass too low for the observed radio luminosity, and X-ray luminosity constraints, if QRS121102 were an AGN. Finally, some dwarf galaxies that host SMBH may be the stripped cores of massive galaxies during the tidal interactions with companion systems. We find no nearby galaxy at the same redshift as the QRS121102 host from low-resolution Keck spectra, or from the PanSTARRS catalog. In conclusion, we find no evidence supporting the hypothesis that the persistent radio source associated with FRB 121102 is an AGN. We instead argue that the inferred size, and the flat radio spectrum, favors a plerion interpretation. We urge continued broadband radio monitoring of QRS121102 to search for long-term evolution, and the detailed evaluation of potential analogs that may provide greater insight into the nature of this class of object.

0 citations0 reviewsNo code linkedOpen access
preprint2022arXiv

Chance-constrained DC Optimal Power Flow with Non-Gaussian Distributed Uncertainties

Chance-constrained programming (CCP) is a promising approach to handle uncertainties in optimal power flow (OPF). However, conventional CCP usually assumes that uncertainties follow Gaussian distributions, which may not match reality. A few papers employed the Gaussian mixture model (GMM) to extend CCP to cases with non-Gaussian uncertainties, but they are only appropriate for cases with uncertainties on the right-hand side but not applicable to DC OPF that containing left-hand side uncertainties. To address this, we develop a tractable GMM-based chance-constrained DC OPF model. In this model, we not only leverage GMM to capture the probability characteristics of non-Gaussian distributed uncertainties, but also develop a linearization technique to reformulate the chance constraints with non-Gaussian distributed uncertainties on the left-hand side into tractable forms. A mathematical proof is further provided to demonstrate that the corresponding reformulation is a safe approximation of the original problem, which guarantees the feasibility of solutions.

0 citations0 reviewsNo code linkedOpen access
preprint2022arXiv

Chance-constrained regulation capacity offering for HVAC systems under non-Gaussian uncertainties with mixture-model-based convexification

Heating, ventilation, and air-conditioning (HVAC) systems are ideal demand-side flexible resources to provide regulation services. However, finding the best hourly regulation capacity offers for HVAC systems in a power market ahead of time is challenging because they are affected by non-Gaussian uncertainties from regulation signals. Moreover, since HVAC systems need to frequently regulate their power according to regulation signals, numerous thermodynamic constraints are introduced, leading to a huge computational burden. This paper proposes a tractable chance-constrained model to address these challenges. It first develops a temporal compression approach, in which the extreme indoor temperatures in the operating hour are estimated and restricted in the comfortable range so that the numerous thermodynamic constraints can be compressed into only a few ones. Then, a novel convexification method is proposed to handle the non-Gaussian uncertainties. This method leverages the Gaussian mixture model to reformulate the chance constraints with non-Gaussian uncertainties on the left-hand side into deterministic non-convex forms. We further prove that these non-convex forms can be approximately convexified by second-order cone constraints with marginal optimality loss. Therefore, the proposed model can be efficiently solved with guaranteed optimality. Numerical experiments are conducted to validate the superiority of the proposed method.

0 citations0 reviewsNo code linkedOpen access
preprint2022arXiv

Deep-quantile-regression-based surrogate model for joint chance-constrained optimal power flow with renewable generation

Joint chance-constrained optimal power flow (JCC-OPF) is a promising tool to manage uncertainties from distributed renewable generation. However, most existing works are based on power flow equations, which require accurate network parameters that may be unobservable in many distribution systems. To address this issue, this paper proposes a learning-based surrogate model for JCC-OPF with renewable generation. This model equivalently converts joint chance constraints in quantile-based forms and introduces deep quantile regression to replicate them, in which a multi-layer perceptron (MLP) is trained with a special loss function to predict the quantile of constraint violations. Another MLP is trained to predict the expected power loss. Then, the JCC-OPF can be formulated without network parameters by reformulating these two MLPs into mixed-integer linear constraints. To further improve its performance, two pre-processing steps, i.e., data augmentation and calibration, are developed. The former trains a simulator to generate more training samples for enhancing the prediction accuracy of MLPs. The latter designs a positive parameter to calibrate the predictions of MLPs so that the feasibility of solutions can be guaranteed. Numerical experiments based on the IEEE 33- and 123-bus systems validate that the proposed model can achieve desirable feasibility and optimality simultaneously with no need for network parameters.

0 citations0 reviewsNo code linkedOpen access
preprint2022arXiv

Efficient constraint learning for data-driven active distribution network operation

Scheduling flexible sources to promote the integration of renewable generation is one fundamental problem for operating active distribution networks (ADNs). However, existing works are usually based on power flow models, which require network parameters (e.g., topology and line impedance) that may be unavailable in practice. To address this issue, we propose an efficient constraint learning method to operate ADNs. This method first trains multilayer perceptrons (MLPs) based on historical data to learn the mappings from decisions to constraint violations and power loss. Then, power flow constraints can be replicated by these MLPs without network parameters. We further prove that MLPs learn constraints by formulating a union of disjoint polytopes to approximate the corresponding feasible region. Thus, the proposed method can be interpreted as a piecewise linearization method, which also explains its desirable ability to replicate complex constraints. Finally, a two-step simplification method is developed to reduce its computational burden. The first step prunes away unnecessary polytopes from the union above. The second step drops redundant linear constraints for each retained polytope. Numerical experiments based on the IEEE 33- and 123-bus test systems validate that the proposed method can achieve desirable optimality and feasibility simultaneously with guaranteed computational efficiency.

0 citations0 reviewsNo code linkedOpen access
preprint2022arXiv

Optimal Control for Unmanned Systems with One-way Broadcast Communication

Unmanned systems (USs) including unmanned aerial vehicles, unmanned underwater vehicles, and unmanned ground vehicles have great application prospects in military and civil fields, among which the process of finding feasible and optimal paths for the agents in USs is a kernel problem. Traditional path finding algorithms are hard to adequately obtain optimal paths in real-time under fast time-varying and poor communication environments. We propose an online optimal control algorithm for USs based on a one-way broadcast communication mode under the assumption of a poor communication environment, mobile targets, radars (or sonar), and missiles (or torpedoes). With the principle of receding horizon control, optimal (or suboptimal) paths are then generated by the approximation theory of neural networks and gradient optimization techniques, with low computation requirements. Also, we give a convergence analysis for our algorithm, and show that each agent can reach its target in finite time under some conditions on agents, targets and radar-missiles. Moreover, simulations demonstrate that the agents in USs can generate optimal (or suboptimal) paths in real time using our algorithm while effectively avoiding collision with other agents or detection by enemy radars.

0 citations0 reviewsNo code linkedOpen access
preprint2020arXiv

Quasi-synchronization of bounded confidence opinion dynamics with stochastic asynchronous rule

Recently the theory of noise-induced synchronization of Hegselmann-Krause (HK) dynamics has been well developed. As a typical opinion dynamics of bounded confidence, the HK model obeys a synchronous updating rule, i.e., \emph{all} agents check and update their opinions at each time point. However, whether asynchronous bounded confidence models, including the famous Deffuant-Weisbuch (DW) model, can be synchronized by noise have not been theoretically proved. In this paper, we propose a generalized bounded confidence model which possesses a stochastic asynchronous rule. The model takes the DW model and the HK model as special cases and can significantly generalize the bounded confidence models to practical application. We discover that the asynchronous model possesses a different noise-based synchronization behavior compared to the synchronous HK model. Generally, the HK dynamics can achieve quasi-synchronization \emph{almost surely} under the drive of noise. For the asynchronous dynamics, we prove that the model can achieve quasi-synchronization \emph{in mean}, which is a new type of quasi-synchronization weaker than the &#34;almost surely&#34; sense. The results unify the theory of noise-induced synchronization of bounded confidence opinion dynamics and hence proves the noise-induced synchronization of DW model theoretically for the first time. Moreover, the results provide a theoretical foundation for developing noise-based control strategy of more complex social opinion systems with stochastic asynchronous rules.

0 citations0 reviewsNo code linkedOpen access
preprint2020arXiv

The multiwavelength counterparts of fast radio bursts

The engines that produce extragalactic fast radio bursts (FRBs), and the mechanism by which the emission is generated, remain unknown. Many FRB models predict prompt multi-wavelength counterparts, which can be used to refine our knowledge of these fundamentals of the FRB phenomenon. However, several previous targeted searches for prompt FRB counterparts have yielded no detections, and have additionally not reached sufficient sensitivity with respect to the predictions. In this work, we demonstrate a technique to estimate the ratio, $η$, between the energy outputs of FRB counterparts at various wavelengths and the radio-wavelength emission. Our technique combines the fluence distribution of the FRB population with results from several wide-field blind surveys for fast transients from the optical to the TeV bands. We present constraints on $η$ that improve upon previous observations even in the case that all unclassified transient events in existing surveys are FRB counterparts. In some scenarios for the FRB engine and emission mechanism, we find that FRB counterparts should have already been detected, thus demonstrating that our technique can successfully test predictions for $η$. However, it is possible that FRB counterparts are lurking amongst catalogs of unclassified transient events. Although our technique is robust to the present uncertainty in the FRB fluence distribution, its ultimate application to accurately estimate or bound $η$ will require the careful analysis of all candidate fast-transient events in multi-wavelength survey data sets.

0 citations0 reviewsNo code linkedOpen access
preprint2017arXiv

EddyNet: A Deep Neural Network For Pixel-Wise Classification of Oceanic Eddies

This work presents EddyNet, a deep learning based architecture for automated eddy detection and classification from Sea Surface Height (SSH) maps provided by the Copernicus Marine and Environment Monitoring Service (CMEMS). EddyNet is a U-Net like network that consists of a convolutional encoder-decoder followed by a pixel-wise classification layer. The output is a map with the same size of the input where pixels have the following labels \{&#39;0&#39;: Non eddy, &#39;1&#39;: anticyclonic eddy, &#39;2&#39;: cyclonic eddy\}. We investigate the use of SELU activation function instead of the classical ReLU+BN and we use an overlap based loss function instead of the cross entropy loss. Keras Python code, the training datasets and EddyNet weights files are open-source and freely available on https://github.com/redouanelg/EddyNet.

0 citations0 reviewsNo code linkedOpen access