Searcharxiv⌕ Search

arXiv subjects

Mu Chen

Publications and source records attributed to Mu Chen.

At least 19 recordsLinked to original sources
arXiv2026

Scaling Multi-Agent Systems with Prospect-State Propagation

Current LLM-based multi-agent systems (MAS) periodically compress intermediate states to reduce inference-time token consumption, thereby attempting to incorporate more agents. However, naive scaling strategies face challenges. For example, in economic simulations, large-scale MAS typically discard semantically rich economic states, i.e., agent behavioral trajectories, which are key drivers of macroeconomic fluctuations. In this paper, we reveal a phenomenon in which agent heterogeneity gradually decreases during simulation, and propose Prospect-State Propagation for Multi-Agent Systems (PspMAS). Inspired by prospect theory, PspMAS decouples each agent's micro state into a compact Prospect State and an expressive Semantic State. The former records psychological traces through a lightweight, parallelizable propagator and continuously injects heterogeneity into the system. The latter leverages the strong perception, reasoning, planning, and decision-making abilities of LLMs. These two components work complementarily, providing a scalable LLM-based multi-agent simulation solution.

cs.MA↗
arXiv2026

Empowering Economic Simulation Through Situation-Aware Llm-Driven Generative System

Traditional economic modeling typically follows a TOP-DOWN paradigm, neglecting individual diversity and the complexity of social interactions. To better capture the complexity of societal structure, Agent-Based Modeling (ABM) employs a BOTTOM-UP solution by incorporating micro-level dynamics to generate macroeconomic phenomena. Reinforcement Learning further improves its decision-making ability through tailored reward signals. However, existing ABM systems struggle to generalize beyond predefined scenarios. Recognizing the potential of LLM-driven role-playing in perception and human-like decision-making, we propose SAMAS, which models individual agents with rich macroeconomic understanding embedded in LLMs and economic trajectories experienced in the passing simulation steps. By jointly modeling both macro-level structural patterns and micro-level dynamic behaviors, SAMAS achieves superior performance in volatility realism and turning point prediction.

cs.MA↗
arXiv2026

A microscopic origin for the breakdown of the Stokes Einstein relation in ion transport

Ion transport underlies the operation of biological ion channels and governs the performance of electrochemical energy-storage devices. A long-standing anomaly is that smaller alkali metal ions, such as Li$^+$, migrate more slowly in water than larger ions, in apparent violation of the Stokes-Einstein relation. This breakdown is conventionally attributed to dielectric friction, a collective drag force arising from electrostatic interactions between a drifting ion and its surrounding solvent. Here, combining nanopore transport measurements over electric fields spanning several orders of magnitude with molecular dynamics simulations, we show that the time-averaged electrostatic force on a migrating ion is not a drag force but a net driving force. By contrasting charged ions with neutral particles, we reveal that ionic charge introduces additional Lorentzian peaks in the frequency-dependent friction coefficient. These peaks originate predominantly from short-range Lennard-Jones (LJ) interactions within the first hydration layer and represent additional channels for energy dissipation, strongest for Li$^+$ and progressively weaker for Na$^+$ and K$^+$. Our results demonstrate that electrostatic interactions primarily act to tighten the local hydration structure, thereby amplifying short-range LJ interactions rather than directly opposing ion motion. This microscopic mechanism provides a unified physical explanation for the breakdown of the Stokes-Einstein relation in aqueous ion transport.

cond-mat.soft↗
arXiv2025

Personalized Chain-of-Thought Summarization of Financial News for Investor Decision Support

Financial advisors and investors struggle with information overload from financial news, where irrelevant content and noise obscure key market signals and hinder timely investment decisions. To address this, we propose a novel Chain-of-Thought (CoT) summarization framework that condenses financial news into concise, event-driven summaries. The framework integrates user-specified keywords to generate personalized outputs, ensuring that only the most relevant contexts are highlighted. These personalized summaries provide an intermediate layer that supports language models in producing investor-focused narratives, bridging the gap between raw news and actionable insights.

q-fin.GN↗
arXiv2025

DIFFVSGG: Diffusion-Driven Online Video Scene Graph Generation

Top-leading solutions for Video Scene Graph Generation (VSGG) typically adopt an offline pipeline. Though demonstrating promising performance, they remain unable to handle real-time video streams and consume large GPU memory. Moreover, these approaches fall short in temporal reasoning, merely aggregating frame-level predictions over a temporal context. In response, we introduce DIFFVSGG, an online VSGG solution that frames this task as an iterative scene graph update problem. Drawing inspiration from Latent Diffusion Models (LDMs) which generate images via denoising a latent feature embedding, we unify the decoding of object classification, bounding box regression, and graph generation three tasks using one shared feature embedding. Then, given an embedding containing unified features of object pairs, we conduct a step-wise Denoising on it within LDMs, so as to deliver a clean embedding which clearly indicates the relationships between objects. This embedding then serves as the input to task-specific heads for object classification, scene graph generation, etc. DIFFVSGG further facilitates continuous temporal reasoning, where predictions for subsequent frames leverage results of past frames as the conditional inputs of LDMs, to guide the reverse diffusion process for current frames. Extensive experiments on three setups of Action Genome demonstrate the superiority of DIFFVSGG.

cs.CV↗
arXiv2025

TGP: Two-modal occupancy prediction with 3D Gaussian and sparse points for 3D Environment Awareness

3D semantic occupancy has rapidly become a research focus in the fields of robotics and autonomous driving environment perception due to its ability to provide more realistic geometric perception and its closer integration with downstream tasks. By performing occupancy prediction of the 3D space in the environment, the ability and robustness of scene understanding can be effectively improved. However, existing occupancy prediction tasks are primarily modeled using voxel or point cloud-based approaches: voxel-based network structures often suffer from the loss of spatial information due to the voxelization process, while point cloud-based methods, although better at retaining spatial location information, face limitations in representing volumetric structural details. To address this issue, we propose a dual-modal prediction method based on 3D Gaussian sets and sparse points, which balances both spatial location and volumetric structural information, achieving higher accuracy in semantic occupancy prediction. Specifically, our method adopts a Transformer-based architecture, taking 3D Gaussian sets, sparse points, and queries as inputs. Through the multi-layer structure of the Transformer, the enhanced queries and 3D Gaussian sets jointly contribute to the semantic occupancy prediction, and an adaptive fusion mechanism integrates the semantic outputs of both modalities to generate the final prediction results. Additionally, to further improve accuracy, we dynamically refine the point cloud at each layer, allowing for more precise location information during occupancy prediction. We conducted experiments on the Occ3DnuScenes dataset, and the experimental results demonstrate superior performance of the proposed method on IoU based metrics.

cs.CV↗
arXiv2024

UAHOI: Uncertainty-aware Robust Interaction Learning for HOI Detection

This paper focuses on Human-Object Interaction (HOI) detection, addressing the challenge of identifying and understanding the interactions between humans and objects within a given image or video frame. Spearheaded by Detection Transformer (DETR), recent developments lead to significant improvements by replacing traditional region proposals by a set of learnable queries. However, despite the powerful representation capabilities provided by Transformers, existing Human-Object Interaction (HOI) detection methods still yield low confidence levels when dealing with complex interactions and are prone to overlooking interactive actions. To address these issues, we propose a novel approach \textsc{UAHOI}, Uncertainty-aware Robust Human-Object Interaction Learning that explicitly estimates prediction uncertainty during the training process to refine both detection and interaction predictions. Our model not only predicts the HOI triplets but also quantifies the uncertainty of these predictions. Specifically, we model this uncertainty through the variance of predictions and incorporate it into the optimization objective, allowing the model to adaptively adjust its confidence threshold based on prediction variance. This integration helps in mitigating the adverse effects of incorrect or ambiguous predictions that are common in traditional methods without any hand-designed components, serving as an automatic confidence threshold. Our method is flexible to existing HOI detection methods and demonstrates improved accuracy. We evaluate \textsc{UAHOI} on two standard benchmarks in the field: V-COCO and HICO-DET, which represent challenging scenarios for HOI detection. Through extensive experiments, we demonstrate that \textsc{UAHOI} achieves significant improvements over existing state-of-the-art methods, enhancing both the accuracy and robustness of HOI detection.

cs.CV↗
arXiv2024

PiPa++: Towards Unification of Domain Adaptive Semantic Segmentation via Self-supervised Learning

Unsupervised domain adaptive segmentation aims to improve the segmentation accuracy of models on target domains without relying on labeled data from those domains. This approach is crucial when labeled target domain data is scarce or unavailable. It seeks to align the feature representations of the source domain (where labeled data is available) and the target domain (where only unlabeled data is present), thus enabling the model to generalize well to the target domain. Current image- and video-level domain adaptation have been addressed using different and specialized frameworks, training strategies and optimizations despite their underlying connections. In this paper, we propose a unified framework PiPa++, which leverages the core idea of ``comparing'' to (1) explicitly encourage learning of discriminative pixel-wise features with intraclass compactness and inter-class separability, (2) promote the robust feature learning of the identical patch against different contexts or fluctuations, and (3) enable the learning of temporal continuity under dynamic environments. With the designed task-smart contrastive sampling strategy, PiPa++ enables the mining of more informative training samples according to the task demand. Extensive experiments demonstrate the effectiveness of our method on both image-level and video-level domain adaption benchmarks. Moreover, the proposed method is compatible with other UDA approaches to further improve the performance without introducing extra parameters.

cs.CV↗
arXiv2024

General and Task-Oriented Video Segmentation

We present GvSeg, a general video segmentation framework for addressing four different video segmentation tasks (i.e., instance, semantic, panoptic, and exemplar-guided) while maintaining an identical architectural design. Currently, there is a trend towards developing general video segmentation solutions that can be applied across multiple tasks. This streamlines research endeavors and simplifies deployment. However, such a highly homogenized framework in current design, where each element maintains uniformity, could overlook the inherent diversity among different tasks and lead to suboptimal performance. To tackle this, GvSeg: i) provides a holistic disentanglement and modeling for segment targets, thoroughly examining them from the perspective of appearance, position, and shape, and on this basis, ii) reformulates the query initialization, matching and sampling strategies in alignment with the task-specific requirement. These architecture-agnostic innovations empower GvSeg to effectively address each unique task by accommodating the specific properties that characterize them. Extensive experiments on seven gold-standard benchmark datasets demonstrate that GvSeg surpasses all existing specialized/general solutions by a significant margin on four different video segmentation tasks.

cs.CV↗
arXiv2024

Quasiparticle scattering in three-dimensional topological insulators near the thickness limit

In the ultra-thin regime, Bi2Te3 films feature two surfaces (with each surface being a two-dimensional Dirac-fermion system) with complicated spin textures and a tunneling term between them. We find in this regime that the quasiparticle scattering is completely different compared with the thick-film case and even behaves differently at each thickness. The thickness-dependent warping effect and tunneling term are found to be the two main factors that govern the scattering behaviors. The inter-band back-scattering that signals the existence of a tunneling term is found to disappear at 4 quintuple layers by the step-edge reflection approach. A four-band model is presented that captures the main features of the thickness-dependent scattering behaviors. Our work clarifies that the prohibition of back-scattering guaranteed by symmetry in topological insulators breaks down in the ultra-thin regime.

cond-mat.mes-hall↗
arXiv2023

Transferring to Real-World Layouts: A Depth-aware Framework for Scene Adaptation

Scene segmentation via unsupervised domain adaptation (UDA) enables the transfer of knowledge acquired from source synthetic data to real-world target data, which largely reduces the need for manual pixel-level annotations in the target domain. To facilitate domain-invariant feature learning, existing methods typically mix data from both the source domain and target domain by simply copying and pasting the pixels. Such vanilla methods are usually sub-optimal since they do not take into account how well the mixed layouts correspond to real-world scenarios. Real-world scenarios are with an inherent layout. We observe that semantic categories, such as sidewalks, buildings, and sky, display relatively consistent depth distributions, and could be clearly distinguished in a depth map. Based on such observation, we propose a depth-aware framework to explicitly leverage depth estimation to mix the categories and facilitate the two complementary tasks, i.e., segmentation and depth learning in an end-to-end manner. In particular, the framework contains a Depth-guided Contextual Filter (DCF) forndata augmentation and a cross-task encoder for contextual learning. DCF simulates the real-world layouts, while the cross-task encoder further adaptively fuses the complementing features between two tasks. Besides, it is worth noting that several public datasets do not provide depth annotation. Therefore, we leverage the off-the-shelf depth estimation network to generate the pseudo depth. Extensive experiments show that our proposed methods, even with pseudo depth, achieve competitive performance on two widely-used bench-marks, i.e. 77.7 mIoU on GTA to Cityscapes and 69.3 mIoU on Synthia to Cityscapes.

cs.CV↗
arXiv2022

PiPa: Pixel- and Patch-wise Self-supervised Learning for Domain Adaptative Semantic Segmentation

Unsupervised Domain Adaptation (UDA) aims to enhance the generalization of the learned model to other domains. The domain-invariant knowledge is transferred from the model trained on labeled source domain, e.g., video game, to unlabeled target domains, e.g., real-world scenarios, saving annotation expenses. Existing UDA methods for semantic segmentation usually focus on minimizing the inter-domain discrepancy of various levels, e.g., pixels, features, and predictions, for extracting domain-invariant knowledge. However, the primary intra-domain knowledge, such as context correlation inside an image, remains underexplored. In an attempt to fill this gap, we propose a unified pixel- and patch-wise self-supervised learning framework, called PiPa, for domain adaptive semantic segmentation that facilitates intra-image pixel-wise correlations and patch-wise semantic consistency against different contexts. The proposed framework exploits the inherent structures of intra-domain images, which: (1) explicitly encourages learning the discriminative pixel-wise features with intra-class compactness and inter-class separability, and (2) motivates the robust feature learning of the identical patch against different contexts or fluctuations. Extensive experiments verify the effectiveness of the proposed method, which obtains competitive accuracy on the two widely-used UDA benchmarks, i.e., 75.6 mIoU on GTA to Cityscapes and 68.2 mIoU on Synthia to Cityscapes. Moreover, our method is compatible with other UDA approaches to further improve the performance without introducing extra parameters.

cs.CV↗
arXiv2022

Selective Trapping of Hexagonally Warped Topological Surface States in a Triangular Quantum Corral

The surface of a three-dimensional topological insulator (TI) hosts two-dimensional massless Dirac fermions (DFs), the gapless and spin-helical nature of which yields many exotic phenomena, such as the immunity of topological surface states (TSS) to back-scattering. This leads to their high transmission through surface defects or potential barriers. Quantum corrals, previously elaborated on metal surfaces, can act as nanometer-sized electronic resonators to trap Schr\"odinger electrons by quantum confinement. It is thus intriguing, concerning their peculiar nature, to put the Dirac electrons of TSS to the test in similar circumstances. Here, we report the behaviors of TSS in a triangular quantum corral (TQC) fabricated by epitaxially growing Bi bilayer nanostructures on the surfaces of Bi2Te3 films. Unlike a circular corral, the TQC is supposed to be totally transparent for DFs. By mapping the electronic structure of TSS inside TQCs through a low-temperature scanning tunneling microscope in the real space, both the trapping and de-trapping behaviors of the TSS electrons are observed. The selection rules are found to be governed by the geometry and spin texture of the constant energy contour of TSS upon the strong hexagonal warping in Bi2Te3. Careful analysis of the quantum interference patterns of quasi-bound states yields the corresponding wave vectors of trapped TSS, through which two trapping mechanisms favoring momenta in different directions are uncovered. Our work indicates the extended nature of TSS and elucidates the selection rules of the trapping of TSS in the presence of a complicated surface state structure, giving insights into the effective engineering of DFs in TIs.

cond-mat.mes-hall↗
arXiv2014

Anisotropic Fabry-P\'{e}rot resonant states confined within nano-steps on the topological insulator surface

The peculiar nature of topological surface states, such as absence of backscattering, weak anti-localization, and quantum anomalous Hall effect, has been demonstrated mainly in bulk and film of topological insulator (TI), using surface sensitive probes and bulk transport probes. However, it is equally important and experimentally challenging to confine massless Dirac fermions with nano-steps on TI surfaces. This potential structure has similar ground with linearly-dispersed photons in Fabry-P\'{e}rot resonators, while reserving fundamental differences from well-studied Fabry-P\'{e}rot resonators and quantum corrals on noble metal surfaces. In this paper, we study the massless Dirac fermions confined within steps along the $x$ ($\Gamma\mathtt{-}$K) or $y$ ($\Gamma\mathtt{-}$M) direction on the TI surface, and the Fabry-P\'{e}rot-like resonances in the electronic local density of states (LDOS) between the steps are found. Due to the remarkable warping effect in the topological surface states, the LDOS confined in the step-well running along $\Gamma$-M direction exhibit anisotropic resonance patterns as compared to those in the step-well along $\Gamma$-K direction, which can be detected by scanning tunneling microscopy. The transmittance properties and spin orientation of Dirac fermion in both cases are also anisotropic in the presence of warping effect.

cond-mat.mes-hall↗
arXiv2013

Direct observation of quantum confinement of massless Dirac fermions in a topological insulator

Since the discovery of topological insulators (TIs)1,2, the peculiar nature of their chiral surface states has been experimentally demonstrated both in bulk and in film materials with open boundaries3,4. Closed boundary on a TI surface may intrigue more interesting phenomena such as quantum confinement of massless Dirac fermions (DFs), which is analogous to the quantum corral (QC) for massive free electrons on a metal surface5-10. To date, it keeps a highly stringent challenge to realize a true Dirac QC due to the unusual transmitting power of a massless fermion. Through heteroepitaxially growing a Bi bilayer on the Bi2Te3 surface with appropriate coverage, here we demonstrate the realization of a true Dirac QC. Specifically, spectacular maps of quantum interference in equilateral triangle-shaped QCs surrounded by Bi bilayers are directly visualized by using a low-temperature scanning tunneling microscope. The present success is ascribed to a perfect orientation matching between the QC boundary and the stationary-phase scattering of massless DFs. In addition, the quasiparticle lifetime of the confined DFs is also systematically measured and analyzed.

cond-mat.mes-hall↗
arXiv2013

Mass acquisition of Dirac fermions in Cr-doped topological insulator Sb2Te3 films

We introduce time-reversal-symmetry breaking by doping Cr atoms into the topmost quintuple layer or into the bulk of Sb2Te3 thin films. We find that even at a high Cr-doping level the Landau level spectrum keeps a good quality, enabling the first demonstration of deviation of the zero-mode Landau level, induced by the acquisition of a mass term in the surface states in the presence of surface or bulk magnetic doping. The magnitude of the mass term in the surface states increases with increasing Cr-doping level. Our observation suggests Cr-doped Sb2Te3 is a promising candidate for the realization of the proposed novel magnetoelectric effects.

cond-mat.mtrl-sci↗
arXiv2011

Landau quantization and the thickness limit of topological insulator thin films of Sb2Te3

We report the experimental observation of Landau quantization of molecular beam epitaxy grown Sb2Te3 thin films by a low-temperature scanning tunneling microscope. Different from all the reported systems, the Landau quantization in Sb2Te3 topological insulator is not sensitive to the intrinsic substitutional defects in the films. As a result, a nearly perfect linear energy dispersion of surface states as 2D massless Dirac fermion system is achieved. We demonstrate that 4 quintuple layers are the thickness limit for Sb2Te3 thin film being a 3D topological insulator. The mechanism of the Landau level broadening is discussed in terms of enhanced quasiparticle lifetime.

cond-mat.mtrl-sci↗