arXiv subjects

Xiaotong Chen

Publications and source records attributed to Xiaotong Chen.

At least 19 recordsLinked to original sources

arXiv2026

Thermal Conductivity and Temperature-Induced Band Gap Renormalization in Crystalline and Amorphous Ga$_2$O$_3$

The lattice thermal conductivity (LTC) and electron-phonon interactions in crystalline and amorphous gallium oxide are herein determined by coupling a machine-learned interatomic potential, namely the moment tensor potential (MTP) model, to first-principles calculations. Crystalline $β$-Ga$_2$O$_3$ exhibits a substantial band gap renormalization (BGR) of $\sim$0.45 eV at 700 K, with $\sim$0.2 eV caused by zero-point BGR. The computed temperature dependence of BGR induced by classical nuclear motion in $β$-Ga$_2$O$_3$ is stronger than that in amorphous Ga$_2$O$_3$, with the difference in BGR reaching $\sim$0.18 eV at 900 K. Thermal transport calculations reveal that the LTC of amorphous Ga$_2$O$_3$ remains near $0.9$ W$\cdot$ m$^{-1}$$\cdot$K$^{-1}$ for temperatures between 300 K and 700 K, which is approximately an order of magnitude lower than that of crystalline $β$-Ga$_2$O$_3$. Overall, the presented framework provides a computationally tractable and reliable route for predicting properties of semiconductors (both crystalline and amorphous) under operating conditions relevant to microelectronics and optoelectronics.

cond-mat.mtrl-sci↗

arXiv2026

Cosmos 3: Omnimodal World Models for Physical AI

We introduce Cosmos 3, a family of omnimodal world models designed to jointly process and generate language, image, video, audio, and action sequences within a unified mixture-of-transformers architecture. By supporting highly flexible input-output configurations, Cosmos 3 seamlessly unifies critical modalities for Physical AI -- effectively subsuming vision-language models, video generators, world simulators, and world-action models into a single framework. Our evaluation demonstrates that Cosmos 3 establishes a new state-of-the-art across a diverse suite of understanding and generation tasks, demonstrating omnimodal world models as scalable, general-purpose backbones for embodied agents. Our post-trained Cosmos 3 models were ranked as the best open-source Text-to-Image and Image-to-Video models by Artificial Analysis, and the best policy model by RoboArena at the time the technical report was written. To accelerate open research and deployment in Physical AI, we make our code, model checkpoints, curated synthetic datasets, and evaluation benchmark available under the Linux Foundation's OpenMDW-1.1 License at https://github.com/nvidia/cosmos and https://huggingface.co/collections/nvidia/cosmos3. The project website is available at https://research.nvidia.com/labs/cosmos-lab/cosmos3.

cs.CV↗

arXiv2026

Lyapunov Stable Graph Neural Flow

Graph Neural Networks (GNNs) are highly vulnerable to adversarial perturbations in both topology and features, making the learning of robust representations a critical challenge. In this work, we bridge GNNs with control theory to introduce a novel defense framework grounded in integer- and fractional-order Lyapunov stability. Unlike conventional strategies that rely on resource-heavy adversarial training or data purification, our approach fundamentally constrains the underlying feature-update dynamics of the GNN. We propose an adaptive, learnable Lyapunov function paired with a novel projection mechanism that maps the network's state into a stable space, thereby offering theoretically provable stability guarantees. Notably, this mechanism is orthogonal to existing defenses, allowing for seamless integration with techniques like adversarial training to achieve cumulative robustness. Extensive experiments demonstrate that our Lyapunov-stable graph neural flows substantially outperform base neural flows and state-of-the-art baselines across standard benchmarks and various adversarial attack scenarios.

cs.LG↗

arXiv2025

Discrete time crystal and perfect many-body tunneling in a periodically driven Heisenberg spin chain

We investigate the non-equilibrium dynamics of a Heisenberg spin-1/2 chain driven by a periodic magnetic field. Based on its instantaneous integrability and inherent symmetry, we analytically study the magnetization and many-body tunneling (MBT). Both of them exhibit periodicity distinct from the driving period. The magnetization is shown to be independent of the initial state and robust against perturbations, signaling the formation of discrete time crystal (DTC) order. The DTC phase is found to be continuously tunable through magnetic field. The system exhibits perfect MBT, manifested as exactly vanishing Loschmidt echo (LE) thus divergent LE rate function at half period of the DTC. Remarkably, the perfect MBT is independent of the system size, and can be traced to an effective gap closure induced by quantum geometric effects. Furthermore, the Loschmidt echo spectra entropy shows logarithmic-dependence on system size, consistent with non-thermal nature of the DTC phase. We propose a protocol using ultracold atoms for experimental realization of the DTC and MBT.

cond-mat.str-el↗

arXiv2025

ACECODER: Acing Coder RL via Automated Test-Case Synthesis

Most progress in recent coder models has been driven by supervised fine-tuning (SFT), while the potential of reinforcement learning (RL) remains largely unexplored, primarily due to the lack of reliable reward data/model in the code domain. In this paper, we address this challenge by leveraging automated large-scale test-case synthesis to enhance code model training. Specifically, we design a pipeline that generates extensive (question, test-cases) pairs from existing code data. Using these test cases, we construct preference pairs based on pass rates over sampled programs to train reward models with Bradley-Terry loss. It shows an average of 10-point improvement for Llama-3.1-8B-Ins and 5-point improvement for Qwen2.5-Coder-7B-Ins through best-of-32 sampling, making the 7B model on par with 236B DeepSeek-V2.5. Furthermore, we conduct reinforcement learning with both reward models and test-case pass rewards, leading to consistent improvements across HumanEval, MBPP, BigCodeBench, and LiveCodeBench (V4). Notably, we follow the R1-style training to start from Qwen2.5-Coder-base directly and show that our RL training can improve model on HumanEval-plus by over 25\% and MBPP-plus by 6\% for merely 80 optimization steps. We believe our results highlight the huge potential of reinforcement learning in coder models.

cs.SE↗

arXiv2025

EditRoom: LLM-parameterized Graph Diffusion for Composable 3D Room Layout Editing

Given the steep learning curve of professional 3D software and the time-consuming process of managing large 3D assets, language-guided 3D scene editing has significant potential in fields such as virtual reality, augmented reality, and gaming. However, recent approaches to language-guided 3D scene editing either require manual interventions or focus only on appearance modifications without supporting comprehensive scene layout changes. In response, we propose EditRoom, a unified framework capable of executing a variety of layout edits through natural language commands, without requiring manual intervention. Specifically, EditRoom leverages Large Language Models (LLMs) for command planning and generates target scenes using a diffusion-based method, enabling six types of edits: rotate, translate, scale, replace, add, and remove. To address the lack of data for language-guided 3D scene editing, we have developed an automatic pipeline to augment existing 3D scene synthesis datasets and introduced EditRoom-DB, a large-scale dataset with 83k editing pairs, for training and evaluation. Our experiments demonstrate that our approach consistently outperforms other baselines across all metrics, indicating higher accuracy and coherence in language-guided scene layout editing.

cs.GR↗

arXiv2025

Static Batching of Irregular Workloads on GPUs: Framework and Application to Efficient MoE Model Inference

It has long been a problem to arrange and execute irregular workloads on massively parallel devices. We propose a general framework for statically batching irregular workloads into a single kernel with a runtime task mapping mechanism on GPUs. We further apply this framework to Mixture-of-Experts (MoE) model inference and implement an optimized and efficient CUDA kernel. Our MoE kernel achieves up to 91% of the peak Tensor Core throughput on NVIDIA H800 GPU and 95% on NVIDIA H20 GPU.

cs.DC↗

arXiv2024

TE2Rules: Explaining Tree Ensembles using Rules

Tree Ensemble (TE) models, such as Gradient Boosted Trees, often achieve optimal performance on tabular datasets, yet their lack of transparency poses challenges for comprehending their decision logic. This paper introduces TE2Rules (Tree Ensemble to Rules), a novel approach for explaining binary classification tree ensemble models through a list of rules, particularly focusing on explaining the minority class. Many state-of-the-art explainers struggle with minority class explanations, making TE2Rules valuable in such cases. The rules generated by TE2Rules closely approximate the original model, ensuring high fidelity, providing an accurate and interpretable means to understand decision-making. Experimental results demonstrate that TE2Rules scales effectively to tree ensembles with hundreds of trees, achieving higher fidelity within runtimes comparable to baselines. TE2Rules allows for a trade-off between runtime and fidelity, enhancing its practical applicability. The implementation is available here: https://github.com/linkedin/TE2Rules.

cs.LG↗

arXiv2023

Tabletop Transparent Scene Reconstruction via Epipolar-Guided Optical Flow with Monocular Depth Completion Prior

Reconstructing transparent objects using affordable RGB-D cameras is a persistent challenge in robotic perception due to inconsistent appearances across views in the RGB domain and inaccurate depth readings in each single-view. We introduce a two-stage pipeline for reconstructing transparent objects tailored for mobile platforms. In the first stage, off-the-shelf monocular object segmentation and depth completion networks are leveraged to predict the depth of transparent objects, furnishing single-view shape prior. Subsequently, we propose Epipolar-guided Optical Flow (EOF) to fuse several single-view shape priors from the first stage to a cross-view consistent 3D reconstruction given camera poses estimated from opaque part of the scene. Our key innovation lies in EOF which employs boundary-sensitive sampling and epipolar-line constraints into optical flow to accurately establish 2D correspondences across multiple views on transparent objects. Quantitative evaluations demonstrate that our pipeline significantly outperforms baseline methods in 3D reconstruction quality, paving the way for more adept robotic perception and interaction with transparent objects.

cs.RO↗

arXiv2023

Valley-polarized Exitonic Mott Insulator in WS2/WSe2 Moiré Superlattice

Strongly enhanced electron-electron interaction in semiconducting moiré superlattices formed by transition metal dichalcogenides (TMDCs) heterobilayers has led to a plethora of intriguing fermionic correlated states. Meanwhile, interlayer excitons in a type-II aligned TMDC heterobilayer moiré superlattice, with electrons and holes separated in different layers, inherit this enhanced interaction and strongly interact with each other, promising for realizing tunable correlated bosonic quasiparticles with valley degree of freedom. We employ photoluminescence spectroscopy to investigate the strong repulsion between interlayer excitons and correlated electrons in a WS2/WSe2 moiré superlattice and combine with theoretical calculations to reveal the spatial extent of interlayer excitons and the band hierarchy of correlated states. We further find that an excitonic Mott insulator state emerges when one interlayer exciton occupies one moiré cell, evidenced by emerging photoluminescence peaks under increased optical excitation power. Double occupancy of excitons in one unit cell requires overcoming the energy cost of exciton-exciton repulsion of about 30-40 meV, depending on the stacking configuration of the WS2/WSe2 heterobilayer. Further, the valley polarization of the excitonic Mott insulator state is enhanced by nearly one order of magnitude. Our study demonstrates the WS2/WSe2 moiré superlattice as a promising platform for engineering and exploring new correlated states of fermion, bosons, and a mixture of both.

cond-mat.mes-hall↗

arXiv2023

Exciton Superposition across Moiré States in a Semiconducting Moiré Superlattice

Moiré superlattices of semiconducting transition metal dichalcogenides (TMDCs) enable unprecedented spatial control of electron wavefunctions in an artificial lattice with periodicities more than ten times larger than that of atomic crystals, leading to emerging quantum states with fascinating electronic and optical properties. The breaking of translational symmetry further introduces a new degree of freedom inside each moiré unit cell: high symmetry points of energy minima called moiré sites, behaving as spatially separated quantum dots. The superposition of a quasiparticle wavefunction between different moiré sites will enable a new platform for quantum information processing but is hindered by the suppressed electron tunneling between moiré sites. Here we demonstrate the superposition between two moiré sites by constructing an angle-aligned trilayer WSe2/monolayer WS2 moiré heterojunction. The two moiré sites with energy minimum allow the formation of two different interlayer excitons, with the hole residing in either moiré site of the first WSe2 layer interfacing the WS2 layer and the electron in the third WSe2 layer. An external electric field can drive the hybridization of either of the interlayer excitons with the intralayer excitons in the third WSe2 layer, realizing the continuous tuning of interlayer exciton hopping between two moiré sites. Therefore, a superposition of the two interlayer excitons localized at different moiré sites can be realized, which can be resolved in the electric-field-dependent optical reflectance spectra, distinctly different from that of the natural trilayer WSe2 in which the moiré modulation is absent. Our study illustrates a strategy of harnessing the new moiré site degree of freedom for quantum information science, a new direction of twistronics.

cond-mat.mes-hall↗

arXiv2023

TransNet: Transparent Object Manipulation Through Category-Level Pose Estimation

Transparent objects present multiple distinct challenges to visual perception systems. First, their lack of distinguishing visual features makes transparent objects harder to detect and localize than opaque objects. Even humans find certain transparent surfaces with little specular reflection or refraction, like glass doors, difficult to perceive. A second challenge is that depth sensors typically used for opaque object perception cannot obtain accurate depth measurements on transparent surfaces due to their unique reflective properties. Stemming from these challenges, we observe that transparent object instances within the same category, such as cups, look more similar to each other than to ordinary opaque objects of that same category. Given this observation, the present paper explores the possibility of category-level transparent object pose estimation rather than instance-level pose estimation. We propose \textit{\textbf{TransNet}}, a two-stage pipeline that estimates category-level transparent object pose using localized depth completion and surface normal estimation. TransNet is evaluated in terms of pose estimation accuracy on a large-scale transparent object dataset and compared to a state-of-the-art category-level pose estimation approach. Results from this comparison demonstrate that TransNet achieves improved pose estimation accuracy on transparent objects. Moreover, we use TransNet to build an autonomous transparent object manipulation system for robotic pick-and-place and pouring tasks.

cs.RO↗

arXiv2023

Large effective magnetic fields from chiral phonons in rare-earth halides

Time-reversal symmetry (TRS) is pivotal for materials optical, magnetic, topological, and transport properties. Chiral phonons, characterized by atoms rotating unidirectionally around their equilibrium positions, generate dynamic lattice structures that break TRS. Here we report that coherent chiral phonons, driven by circularly polarized terahertz light pulses, can polarize the paramagnetic spins in CeF3 like a quasi-static magnetic field on the order of 1 Tesla. Through time-resolved Faraday rotation and Kerr ellipticity, we found the transient magnetization is only excited by pulses resonant with phonons, proportional to the angular momentum of the phonons, and growing with magnetic susceptibility at cryogenic temperatures, as expected from the spin-phonon coupling model. The time-dependent effective magnetic field quantitatively agrees with that calculated from phonon dynamics. Our results may open a new route to directly investigate mode-specific spin-phonon interaction in ultrafast magnetism, energy-efficient spintronics, and non-equilibrium phases of matter with broken TRS.

cond-mat.mtrl-sci↗

arXiv2023

An Operational Perspective to Fairness Interventions: Where and How to Intervene

As AI-based decision systems proliferate, their successful operationalization requires balancing multiple desiderata: predictive performance, disparity across groups, safeguarding sensitive group attributes (e.g., race), and engineering cost. We present a holistic framework for evaluating and contextualizing fairness interventions with respect to the above desiderata. The two key points of practical consideration are \emph{where} (pre-, in-, post-processing) and \emph{how} (in what way the sensitive group data is used) the intervention is introduced. We demonstrate our framework with a case study on predictive parity. In it, we first propose a novel method for achieving predictive parity fairness without using group data at inference time via distibutionally robust optimization. Then, we showcase the effectiveness of these methods in a benchmarking study of close to 400 variations across two major model types (XGBoost vs. Neural Net), ten datasets, and over twenty unique methodologies. Methodological insights derived from our empirical study inform the practical design of ML workflow with fairness as a central concern. We find predictive parity is difficult to achieve without using group data, and despite requiring group data during model training (but not inference), distributionally robust methods we develop provide significant Pareto improvement. Moreover, a plain XGBoost model often Pareto-dominates neural networks with fairness interventions, highlighting the importance of model inductive bias.

cs.LG↗

arXiv2022

Pair DETR: Contrastive Learning Speeds Up DETR Training

The DETR object detection approach applies the transformer encoder and decoder architecture to detect objects and achieves promising performance. In this paper, we present a simple approach to address the main problem of DETR, the slow convergence, by using representation learning technique. In this approach, we detect an object bounding box as a pair of keypoints, the top-left corner and the center, using two decoders. By detecting objects as paired keypoints, the model builds up a joint classification and pair association on the output queries from two decoders. For the pair association we propose utilizing contrastive self-supervised learning algorithm without requiring specialized architecture. Experimental results on MS COCO dataset show that Pair DETR can converge at least 10x faster than original DETR and 1.5x faster than Conditional DETR during training, while having consistently higher Average Precision scores.

cs.CV↗

arXiv2022

Manipulation-Oriented Object Perception in Clutter through Affordance Coordinate Frames

In order to enable robust operation in unstructured environments, robots should be able to generalize manipulation actions to novel object instances. For example, to pour and serve a drink, a robot should be able to recognize novel containers which afford the task. Most importantly, robots should be able to manipulate these novel containers to fulfill the task. To achieve this, we aim to provide robust and generalized perception of object affordances and their associated manipulation poses for reliable manipulation. In this work, we combine the notions of affordance and category-level pose, and introduce the Affordance Coordinate Frame (ACF). With ACF, we represent each object class in terms of individual affordance parts and the compatibility between them, where each part is associated with a part category-level pose for robot manipulation. In our experiments, we demonstrate that ACF outperforms state-of-the-art methods for object detection, as well as category-level pose estimation for object parts. We further demonstrate the applicability of ACF to robot manipulation tasks through experiments in a simulated environment.

cs.RO↗

arXiv2022

TransNet: Category-Level Transparent Object Pose Estimation

Transparent objects present multiple distinct challenges to visual perception systems. First, their lack of distinguishing visual features makes transparent objects harder to detect and localize than opaque objects. Even humans find certain transparent surfaces with little specular reflection or refraction, e.g. glass doors, difficult to perceive. A second challenge is that common depth sensors typically used for opaque object perception cannot obtain accurate depth measurements on transparent objects due to their unique reflective properties. Stemming from these challenges, we observe that transparent object instances within the same category (e.g. cups) look more similar to each other than to ordinary opaque objects of that same category. Given this observation, the present paper sets out to explore the possibility of category-level transparent object pose estimation rather than instance-level pose estimation. We propose TransNet, a two-stage pipeline that learns to estimate category-level transparent object pose using localized depth completion and surface normal estimation. TransNet is evaluated in terms of pose estimation accuracy on a recent, large-scale transparent object dataset and compared to a state-of-the-art category-level pose estimation approach. Results from this comparison demonstrate that TransNet achieves improved pose estimation accuracy on transparent objects and key findings from the included ablation studies suggest future directions for performance improvements.

cs.CV↗

arXiv2022

VLMbench: A Compositional Benchmark for Vision-and-Language Manipulation

Benefiting from language flexibility and compositionality, humans naturally intend to use language to command an embodied agent for complex tasks such as navigation and object manipulation. In this work, we aim to fill the blank of the last mile of embodied agents -- object manipulation by following human guidance, e.g., "move the red mug next to the box while keeping it upright." To this end, we introduce an Automatic Manipulation Solver (AMSolver) system and build a Vision-and-Language Manipulation benchmark (VLMbench) based on it, containing various language instructions on categorized robotic manipulation tasks. Specifically, modular rule-based task templates are created to automatically generate robot demonstrations with language instructions, consisting of diverse object shapes and appearances, action types, and motion constraints. We also develop a keypoint-based model 6D-CLIPort to deal with multi-view observations and language input and output a sequence of 6 degrees of freedom (DoF) actions. We hope the new simulator and benchmark will facilitate future research on language-guided robotic manipulation.

cs.RO↗