arXiv subjects

Yanzhen Wang

Publications and source records attributed to Yanzhen Wang.

8 recordsLinked to original sources

arXiv2026

Reframing Long-Tailed Learning via Loss Landscape Geometry

Balancing performance trade-off on long-tail (LT) data distributions remains a long-standing challenge. In this paper, we posit that this dilemma stems from a phenomenon called "tail performance degradation" (the model tends to severely overfit on head classes while quickly forgetting tail classes) and pose a solution from a loss landscape perspective. We observe that different classes possess divergent convergence points in the loss landscape. Besides, this divergence is aggravated when the model settles into sharp and non-robust minima, rather than a shared and flat solution that is beneficial for all classes. In light of this, we propose a continual learning inspired framework to prevent "tail performance degradation". To avoid inefficient per-class parameter preservation, a Grouped Knowledge Preservation module is proposed to memorize group-specific convergence parameters, promoting convergence towards a shared solution. Concurrently, our framework integrates a Grouped Sharpness Aware module to seek flatter minima by explicitly addressing the geometry of the loss landscape. Notably, our framework requires neither external training samples nor pre-trained models, facilitating the broad applicability. Extensive experiments on four benchmarks demonstrate significant performance gains over state-of-the-art methods. The code is available at:https://gkp-gsa.github.io/.

cs.CV↗

arXiv2026

DeepH-pack: A general-purpose neural network package for deep-learning electronic structure calculations

In computational physics and materials science, first-principles methods, particularly density functional theory, have become central tools for electronic structure prediction and materials design. Recently, rapid advances in artificial intelligence (AI) have begun to reshape the research landscape, giving rise to the emerging field of deep-learning electronic structure calculations. Despite numerous pioneering studies, the field remains in its early stages; existing software implementations are often fragmented, lacking unified frameworks and standardized interfaces required for broad community adoption. Here we present DeepH-pack, a comprehensive and unified software package that integrates first-principles calculations with deep learning. By incorporating fundamental physical principles into neural-network design, such as the nearsightedness principle and the equivariance principle, DeepH-pack achieves robust cross-scale and cross-material generalizability. This allows models trained on small-scale structures to generalize to large-scale and previously unseen materials. The toolkit preserves first-principles accuracy while accelerating electronic structure calculations by several orders of magnitude, establishing an efficient and intelligent computational paradigm for large-scale materials simulation, high-throughput materials database construction, and AI-driven materials discovery.

cond-mat.mtrl-sci↗

arXiv2025

QMBench: A Research Level Benchmark for Quantum Materials Research

We introduce QMBench, a comprehensive benchmark designed to evaluate the capability of large language model agents in quantum materials research. This specialized benchmark assesses the model's ability to apply condensed matter physics knowledge and computational techniques such as density functional theory to solve research problems in quantum materials science. QMBench encompasses different domains of the quantum material research, including structural properties, electronic properties, thermodynamic and other properties, symmetry principle and computational methodologies. By providing a standardized evaluation framework, QMBench aims to accelerate the development of an AI scientist capable of making creative contributions to quantum materials research. We expect QMBench to be developed and constantly improved by the research community.

cond-mat.mtrl-sci↗

arXiv2024

Visualizing incommensurate inter-valley coherent states in rhombohedral trilayer graphene

ABC-stacked rhombohedral graphene multilayers exhibit a wide variety of electronic ground states characterized by broken isospin symmetry and superconductivity. Recently, indirect evidence of inter-valley coherent (IVC) order has been reported in rhombohedral trilayer graphene (RTG), with possible implications for the origin of superconductivity. Here, we report the direct visualization of IVC order in RTG using scanning tunneling microscopy and spectroscopy. Tuning the chemical potential through the Van Hove singularity near the edge of the valence band, we observe a cascade of phase transitions associated with the formation of half- and quarter-metal states. IVC phases, distinguished by an enlarged real space unit cell, are directly imaged near both the high- and low-density boundaries of the half-metal phase. At high hole density, we precisely reconstruct the IVC band structure through quasiparticle interference. Intriguingly, the charge density modulations reveal a C3-symmetric incommensurate IVC order that agrees with the recent prediction of an IVC-crystal phase. Our findings demonstrate that IVC phases are a widespread symmetry-broken ground state within graphene systems.

cond-mat.mes-hall↗

arXiv2024

DeepH-2: Enhancing deep-learning electronic structure via an equivariant local-coordinate transformer

Deep-learning electronic structure calculations show great potential for revolutionizing the landscape of computational materials research. However, current neural-network architectures are not deemed suitable for widespread general-purpose application. Here we introduce a framework of equivariant local-coordinate transformer, designed to enhance the deep-learning density functional theory Hamiltonian referred to as DeepH-2. Unlike previous models such as DeepH and DeepH-E3, DeepH-2 seamlessly integrates the simplicity of local-coordinate transformations and the mathematical elegance of equivariant neural networks, effectively overcoming their respective disadvantages. Based on our comprehensive experiments, DeepH-2 demonstrates superiority over its predecessors in both efficiency and accuracy, showcasing state-of-the-art performance. This advancement opens up opportunities for exploring universal neural network models or even large materials models.

physics.comp-ph↗

arXiv2024

A Study on Training and Developing Large Language Models for Behavior Tree Generation

This paper presents an innovative exploration of the application potential of large language models (LLM) in addressing the challenging task of automatically generating behavior trees (BTs) for complex tasks. The conventional manual BT generation method is inefficient and heavily reliant on domain expertise. On the other hand, existing automatic BT generation technologies encounter bottlenecks related to task complexity, model adaptability, and reliability. In order to overcome these challenges, we propose a novel methodology that leverages the robust representation and reasoning abilities of LLMs. The core contribution of this paper lies in the design of a BT generation framework based on LLM, which encompasses the entire process, from data synthesis and model training to application developing and data verification. Synthetic data is introduced to train the BT generation model (BTGen model), enhancing its understanding and adaptability to various complex tasks, thereby significantly improving its overall performance. In order to ensure the effectiveness and executability of the generated BTs, we emphasize the importance of data verification and introduce a multilevel verification strategy. Additionally, we explore a range of agent design and development schemes with LLM as the central element. We hope that the work in this paper may provide a reference for the researchers who are interested in BT generation based on LLMs.

cs.CL↗

arXiv2023

Spin Squeezing through Collective Spin-Spin Interactions

Spin squeezing provides crucial quantum resource for quantum metrology and quantum information science. Here we propose that one axis-twisted (OAT) spin squeezing can be generated from free evolution under a general coupled-spin model with collective spin-spin interactions. We further propose pulse schemes to recover squeezing from parameter imperfections, and reach the extreme squeezing with Heisenberg-limited measurement precision scaling as $1/N$ for $N$ particles. This work provides a feasible method for generating extreme spin squeezing.

quant-ph↗

arXiv2023

Heisenberg-limited spin squeezing in coupled spin systems

Spin squeezing plays a crucial role in quantum metrology and quantum information science. Its generation is the prerequisite for further applications but still faces an enormous challenge since the existing physical systems rarely contain the required squeezing interactions. Here we propose a universal scheme to generate spin squeezing in coupled spin models with collective spin-spin interactions, which commonly exist in various systems. Our scheme can transform the coupled spin interactions into squeezing interactions, and reach the extreme squeezing with Heisenberg-limited measurement precision scaling as $1/N$ for $N$ particles. Only constant and continuous driving fields are required, which is accessible to a series of current realistic experiments. This work greatly enriches the variety of systems that can generate the Heisenberg-limited spin squeezing, with broad applications in quantum precision measurement.

quant-ph↗