arXiv subjects

Haichen Wang

Publications and source records attributed to Haichen Wang.

12 recordsLinked to original sources

arXiv2026

Optical selection rules in hexagonal Ge polytypes and their lifting by symmetry perturbation

Hexagonal germanium polytypes have emerged as promising direct-gap semiconductors for silicon-integrated optoelectronics, yet their optical properties remain largely unexplored beyond the well-studied 2H phase. We present a comprehensive theoretical study of optical properties of hexagonal 2H-, 4H-, and 6H-Ge polytypes through ab initio calculations of quasiparticle band structures, dipole transition matrix elements, and solution of the Bethe-Salpeter equation. While all three polytypes exhibit direct band gaps of increasing size from 2H to 6H, we reveal that the fundamental optical transition in 4H-Ge is parity-forbidden due to matching band parities at the valence and conduction band edges. This selection rule results in a radiative lifetime seven orders of magnitude longer than in 2H- and 6H-Ge, severely limiting light emission capabilities. To demonstrate that the selection rule can be lifted, we introduce controlled symmetry perturbations by substituting single Ge atoms with Si in each unit cell, breaking the crystal symmetry. This perturbation increases the optical matrix elements by up to two orders of magnitude and reduces radiative lifetimes for all perturbed polytypes. We also compute absorption coefficients and frequency-dependent dielectric tensors for both light polarizations, including excitonic effects up to 5 eV, providing complete optical characterization of ideal and symmetry-perturbed hexagonal Ge systems relevant for optoelectronic applications.

cond-mat.mtrl-sci↗

arXiv2025

Automating High Energy Physics Data Analysis with LLM-Powered Agents

We present a proof-of-principle study demonstrating the use of large language model (LLM) agents to automate a representative high energy physics (HEP) analysis. Using the Higgs boson diphoton cross-section measurement as a case study with ATLAS Open Data, we design a hybrid system that combines an LLM-based supervisor-coder agent with the Snakemake workflow manager. In this architecture, the workflow manager enforces reproducibility and determinism, while the agent autonomously generates, executes, and iteratively corrects analysis code in response to user instructions. We define quantitative evaluation metrics including success rate, error distribution, costs per specific task, and average number of API calls, to assess agent performance across multi-stage workflows. To characterize variability across architectures, we benchmark a representative selection of state-of-the-art LLMs spanning the Gemini and GPT-5 series, the Claude family, and leading open-weight models. While the workflow manager ensures deterministic execution of all analysis steps, the final outputs still show stochastic variation. Although we set the temperature to zero, other sampling parameters (e.g., top-p, top-k) remained at their defaults, and some reasoning-oriented models internally adjust these settings. Consequently, the models do not produce fully deterministic results. This study establishes the first LLM-agent-driven automated data-analysis framework in HEP, enabling systematic benchmarking of model capabilities, stability, and limitations in real-world scientific computing environments. The baseline code used in this work is available at https://huggingface.co/HWresearch/LLM4HEP. This work was accepted as a poster at the Machine Learning and the Physical Sciences (ML4PS) workshop at NeurIPS 2025. The initial submission was made on August 30, 2025.

physics.data-an↗

arXiv2025

Reconstruction of Differentially Private Text Sanitization via Large Language Models

Differential privacy (DP) is the de facto privacy standard against privacy leakage attacks, including many recently discovered ones against large language models (LLMs). However, we discovered that LLMs could reconstruct the altered/removed privacy from given DP-sanitized prompts. We propose two attacks (black-box and white-box) based on the accessibility to LLMs and show that LLMs could connect the pair of DP-sanitized text and the corresponding private training data of LLMs by giving sample text pairs as instructions (in the black-box attacks) or fine-tuning data (in the white-box attacks). To illustrate our findings, we conduct comprehensive experiments on modern LLMs (e.g., LLaMA-2, LLaMA-3, ChatGPT-3.5, ChatGPT-4, ChatGPT-4o, Claude-3, Claude-3.5, OPT, GPT-Neo, GPT-J, Gemma-2, and Pythia) using commonly used datasets (such as WikiMIA, Pile-CC, and Pile-Wiki) against both word-level and sentence-level DP. The experimental results show promising recovery rates, e.g., the black-box attacks against the word-level DP over WikiMIA dataset gave 72.18% on LLaMA-2 (70B), 82.39% on LLaMA-3 (70B), 75.35% on Gemma-2, 91.2% on ChatGPT-4o, and 94.01% on Claude-3.5 (Sonnet). More urgently, this study indicates that these well-known LLMs have emerged as a new security risk for existing DP text sanitization approaches in the current environment.

cs.CR↗

arXiv2025

ADFormer: Aggregation Differential Transformer for Passenger Demand Forecasting

Passenger demand forecasting helps optimize vehicle scheduling, thereby improving urban efficiency. Recently, attention-based methods have been used to adequately capture the dynamic nature of spatio-temporal data. However, existing methods that rely on heuristic masking strategies cannot fully adapt to the complex spatio-temporal correlations, hindering the model from focusing on the right context. These works also overlook the high-level correlations that exist in the real world. Effectively integrating these high-level correlations with the original correlations is crucial. To fill this gap, we propose the Aggregation Differential Transformer (ADFormer), which offers new insights to demand forecasting promotion. Specifically, we utilize Differential Attention to capture the original spatial correlations and achieve attention denoising. Meanwhile, we design distinct aggregation strategies based on the nature of space and time. Then, the original correlations are unified with the high-level correlations, enabling the model to capture holistic spatio-temporal relations. Experiments conducted on taxi and bike datasets confirm the effectiveness and efficiency of our model, demonstrating its practical value. The code is available at https://github.com/decisionintelligence/ADFormer.

cs.AI↗

arXiv2025

Transforming Simulation to Data Without Pairing

We explore a generative machine learning-based approach for estimating multi-dimensional probability density functions (PDFs) in a target sample using a statistically independent but related control sample - a common challenge in particle physics data analysis. The generative model must accurately reproduce individual observable distributions while preserving the correlations between them, based on the input multidimensional distribution from the control sample. Here we present a conditional normalizing flow model (CNF) based on a chain of bijectors which learns to transform unpaired simulation events to data events. We assess the performance of the CNF model in the context of LHC Higgs to diphoton analysis, where we use the CNF model to convert a Monte Carlo diphoton sample to one that models data. We show that the CNF model can accurately model complex data distributions and correlations. We also leverage the recently popularized Modified Differential Multiplier Method (MDMM) to improve the convergence of our model and assign physical meaning to usually arbitrary loss-function parameters.

physics.data-an↗

arXiv2025

A non-orthogonal representation of the chemical space

We present a novel approach to generate a fingerprint for crystalline materials that balances efficiency for machine processing and human interpretability, allowing its application in both machine learning inference and understanding of structure-property relationships. Our proposed material encoding has two components: one representing the crystal structure and the other characterizing the chemical composition, that we call Pettifor embedding. For the latter we construct a non-orthogonal space where each axis represents a chemical element and where the angle between the axes quantifies a measure of the similarity between them. The chemical composition is then defined by the point on the unit sphere in this non-orthogonal space. We show that the Pettifor embeddings systematically outperform other commonly used elemental embeddings in compositional machine learning models. Using the Pettifor embeddings to define a distance metric and applying dimension reduction techniques, we construct a two-dimensional global map of the space of thermodynamically stable crystalline compounds. Despite their simplicity, such maps succeed in providing a physical separation of material classes according to basic physical properties.

cond-mat.mtrl-sci↗

arXiv2024

Pretrained Event Classification Model for High Energy Physics Analysis

We introduce a foundation model for event classification in high-energy physics, built on a Graph Neural Network architecture and trained on 120 million simulated proton-proton collision events spanning 12 distinct physics processes. The model is pretrained to learn a general and robust representation of collision data using challenging multiclass and multilabel classification tasks. Its performance is evaluated across seven event classification tasks, which include new physics processes not encountered during pretraining as well as ATLAS Open Data to demonstrate generalizability across different simulation frameworks, from Delphes fast simulation to full ATLAS detector simulation. Fine-tuning the pretrained model significantly improves classification performance, particularly in scenarios with limited training data, demonstrating gains in both accuracy and computational efficiency. To investigate the underlying mechanisms behind these performance improvements, we employ a representational similarity evaluation framework based on Centered Kernel Alignment. This analysis reveals that encoder-stage representations of the fine-tuned model remain similar to those of the baseline, while intermediate graph processing layers diverge substantially, indicating that fine-tuning preserves general-purpose encoders while developing fundamentally different message-passing pathways to arrive at superior task performance.

hep-ph↗

arXiv2024

Parton Labeling without Matching: Unveiling Emergent Labelling Capabilities in Regression Models

Parton labeling methods are widely used when reconstructing collider events with top quarks or other massive particles. State-of-the-art techniques are based on machine learning and require training data with events that have been matched using simulations with truth information. In nature, there is no unique matching between partons and final state objects due to the properties of the strong force and due to acceptance effects. We propose a new approach to parton labeling that circumvents these challenges by recycling regression models. The final state objects that are most relevant for a regression model to predict the properties of a particular top quark are assigned to said parent particle without having any parton-matched training data. This approach is demonstrated using simulated events with top quarks and outperforms the widely-used $χ^2$ method.

hep-ph↗

arXiv2023

Generative Machine Learning for Detector Response Modeling with a Conditional Normalizing Flow

In this paper, we explore the potential of generative machine learning models as an alternative to the computationally expensive Monte Carlo (MC) simulations commonly used by the Large Hadron Collider (LHC) experiments. Our objective is to develop a generative model capable of efficiently simulating detector responses for specific particle observables, focusing on the correlations between detector responses of different particles in the same event and accommodating asymmetric detector responses. We present a conditional normalizing flow model (CNF) based on a chain of Masked Autoregressive Flows, which effectively incorporates conditional variables and models high-dimensional density distributions. We assess the performance of the \cnf model using a simulated sample of Higgs boson decaying to diphoton events at the LHC. We create reconstruction-level observables using a smearing technique. We show that conditional normalizing flows can accurately model complex detector responses and their correlation. This method can potentially reduce the computational burden associated with generating large numbers of simulated events while ensuring that the generated events meet the requirements for data analyses.

hep-ex↗

arXiv2023

A Holistic Approach to Predicting Top Quark Kinematic Properties with the Covariant Particle Transformer

Precise reconstruction of top quark properties is a challenging task at the Large Hadron Collider due to combinatorial backgrounds and missing information. We introduce a physics-informed neural network architecture called the Covariant Particle Transformer (CPT) for directly predicting the top quark kinematic properties from reconstructed final state objects. This approach is permutation invariant and partially Lorentz covariant and can account for a variable number of input objects. In contrast to previous machine learning-based reconstruction methods, CPT is able to predict top quark four-momenta regardless of the jet multiplicity in the event. Using simulations, we show that the CPT performs favorably compared with other machine learning top quark reconstruction approaches.

hep-ph↗

arXiv2022

Machine Learning guided high-throughput search of non-oxide garnets

Garnets, known since the early stages of human civilization, have found important applications in modern technologies including magnetorestriction, spintronics, lithium batteries, etc. The overwhelming majority of experimentally known garnets are oxides, while explorations (experimental or theoretical) for the rest of the chemical space have been limited in scope. A key issue is that the garnet structure has a large primitive unit cell, requiring an enormous amount of computational resources. To perform a comprehensive search of the complete chemical space for new garnets,we combine recent progress in graph neural networks with high-throughput calculations. We apply the machine learning model to identify the potential (meta-)stable garnet systems before systematic density-functional calculations to validate the predictions. In this way, we discover more than 600 ternary garnets with distances to the convex hull below 100~meV/atom with a variety of physical and chemical properties. This includes sulfide, nitride and halide garnets. For these, we analyze the electronic structure and discuss the connection between the value of the electronic band gap and charge balance.

cond-mat.mtrl-sci↗

arXiv2014

Higgs-flavon mixing and LHC phenomenology in a simplified model of broken flavor symmetry

The LHC phenomenology of a low-scale gauged flavor symmetry model with inverted hierarchy is studied, through introduction of a simplified model of broken flavor symmetry. A new scalar (a flavon) and a new neutral top-philic massive gauge boson emerge with mass in the TeV range along with a new heavy fermion associated with the standard model top quark. After checking constraints from electroweak precision observables, we investigate the influence of the model on Higgs boson physics, notably on its production cross section and decay branching fractions. Limits on the flavon $φ$ from heavy Higgs boson searches at the LHC at 7 and 8 TeV are presented. The branching fractions of the flavon are computed as a function of the flavon mass and the Higgs-flavon mixing angle. We also explore possible discovery of the flavon at 14 TeV, particularly via the $φ\rightarrow Z^0Z^0$ decay channel in the $2\ell2\ell'$ final state, and through standard model Higgs boson pair production $φ\rightarrow hh$ in the $b\bar{b}γγ$ final state. We conclude that the flavon mass range up to $500$ GeV could probed down to quite small values of the Higgs-flavon mixing angle with 100 fb$^{-1}$ of integrated luminosity at 14 TeV.

hep-ph↗