arXiv subjects

Jiahao Su

Publications and source records attributed to Jiahao Su.

At least 19 recordsLinked to original sources

arXiv2026

Development of A Novel Compton Camera for MeV Gamma-Ray Measurement in Space

The astrophysical gamma rays in the MeV energy region have not yet been well-explored due to the limitation of detection technology in the past decades, and the famous gamma-ray "MeV gap" exists. Opening the window of MeV gamma-ray is not only critical for the gamma astronomy but also essential for rich frontier researches in astro-particle physics, such as detecting light dark matter, probing the primordial black hole and better understanding of nucleosynthesis. As a pilot experiment of the project for dark matter detection in space at Shanghai Jiao Tong University, a three-layer Compton camera with the energy resolution better than 4% and position resolution of ~2 mm is developed utilizing the novel scintillators. Here we show the design, detailed calibration and validation results of the novel Compton camera, and demonstrate its good ability of MeV gamma-ray source imaging for the upcoming in-orbit mission.

hep-ex↗

arXiv2026

DV-World: Benchmarking Data Visualization Agents in Real-World Scenarios

Real-world data visualization (DV) requires native environmental grounding, cross-platform evolution, and proactive intent alignment. Yet, existing benchmarks often suffer from code-sandbox confinement, single-language creation-only tasks, and assumption of perfect intent. To bridge these gaps, we introduce DV-World, a benchmark of 260 tasks designed to evaluate DV agents across real-world professional lifecycles. DV-World spans three domains: DV-Sheet for native spreadsheet manipulation including chart and dashboard creation as well as diagnostic repair; DV-Evolution for adapting and restructuring reference visual artifacts to fit new data across diverse programming paradigms and DV-Interact for proactive intent alignment with a user simulator that mimics real-world ambiguous requirements. Our hybrid evaluation framework integrates Table-value Alignment for numerical precision and MLLM-as-a-Judge with rubrics for semantic-visual assessment. Experiments reveal that state-of-the-art models achieve less than 50% overall performance, exposing critical deficits in handling the complex challenges of real-world data visualization. DV-World provides a realistic testbed to steer development toward the versatile expertise required in enterprise workflows. Our data and code are available at \href{https://github.com/DA-Open/DV-World}{this project page}.

cs.CL↗

arXiv2025

Control of localized states of itinerant electrons and their magnetic interactions

Controlling the magnetic properties of nanosystems by an electric field offers a number of advantages for spintronics applications. Using the noncollinear Alexander-Anderson model, we have shown that the interaction of localized magnetic moments formed by itinerant electrons strongly depends on the position of the d-level relative to the Fermi level, which determines the number of localized electrons. Depending on this parameter, the ground state of the magnetic dimer can be ferromagnetic, antiferromagnetic, or noncollinear without the effects of spin-orbit interaction. The magnetic state can be controlled by shifting the d-level with an electric field, even without current flow. For a sufficiently large value of the hopping parameter between localized states there can be several self-consistent solutions with different values of magnetic moments. This opens new possibilities for manipulation of the magnetic structure of nanosystems. The results obtained lead to a new interpretation of the mechanisms of magnetization reversal, recording, and deleting of magnetic structures in tunneling spectroscopy experiments.

physics.comp-ph↗

arXiv2025

Talk2Image: A Multi-Agent System for Multi-Turn Image Generation and Editing

Text-to-image generation tasks have driven remarkable advances in diverse media applications, yet most focus on single-turn scenarios and struggle with iterative, multi-turn creative tasks. Recent dialogue-based systems attempt to bridge this gap, but their single-agent, sequential paradigm often causes intention drift and incoherent edits. To address these limitations, we present Talk2Image, a novel multi-agent system for interactive image generation and editing in multi-turn dialogue scenarios. Our approach integrates three key components: intention parsing from dialogue history, task decomposition and collaborative execution across specialized agents, and feedback-driven refinement based on a multi-view evaluation mechanism. Talk2Image enables step-by-step alignment with user intention and consistent image editing. Experiments demonstrate that Talk2Image outperforms existing baselines in controllability, coherence, and user satisfaction across iterative image generation and editing tasks.

cs.CV↗

arXiv2025

Crossover in the Ordered Phase in the Non-Mermin-Wagner-Hohenberg Regime of Spin Models with Long-Range Coupling

Continuous spin models with long-range interactions of the form $r^{-\sigma}$, where $r$ is the distance between two spins and $\sigma$ controls the decay of the interaction, exhibit enhanced order that competes with thermal disturbances, leading to a richer variety of phases and types of phase transitions. In-depth research in this area not only aids in comprehending the complex behaviors in theoretical models but also provides valuable insights into the diverse phase transitions observed in real materials. Here, we identify that the true long-range ordered phase encompasses distinct scaling regimes, which we term Enhance Long-Range Ordered (EnLRO) and Reduce Long-Range Ordered (ReLRO) regimes. In the one-dimensional XY model, the crossover from EnLRO to ReLRO regimes occurs around $\sigma \approx 1.575$, while in two dimensions, the crossover happens near $\sigma \approx 3.2$. Applying finite-size scaling analysis, we extract the critical exponents that characterize the order-to-disorder phase transitions in the EnLRO and ReLRO regimes, constructing comprehensive phase diagrams. The analysis is further extended to the 1D and 2D long-range Heisenberg models, where we find the EnLRO-ReLRO crossover at $\sigma \approx 1.575$ and $\sigma \approx 3.22$, respectively. The similar crossover points suggest that the distinction between EnLRO and ReLRO regimes is a generic feature in continuous spin models with long-range interactions.

cond-mat.stat-mech↗

arXiv2025

Variation of Bose surface by Filling in Cooper pair Bose metal

The Cooper pair Bose metal (CPBM) is a non-superfluid quantum phase in which uncondensed fermion pairs form a "Bose surface" in momentum space. We investigate the CPBM in the two-dimensional spin-anisotropic attractive Hubbard model by tuning the next-nearest-neighbor (NNN) hopping t', carrier filling n, and spin anisotropy alpha, using large-scale constrained-path quantum Monte Carlo simulations. A moderate NNN hopping (t'/t = 0.2) substantially enlarges the CPBM region: the phase extends into weaker anisotropy regimes and coexists with a commensurate charge-density wave (CDW) near half-filling (n > 0.95), where CDW order would otherwise dominate at t' = 0. Interestingly, t' suppresses the overall CDW peak amplitude and introduces a geometric correlation between the orientations of the Fermi and Bose surfaces: for weak Fermi-surface rotations, the Bose surface remains aligned with the lattice axes, while larger distortions drive both surfaces to rotate in tandem. Momentum-resolved pairing distributions reveal that the bosonic pairing channels are jointly controlled by t' and carrier filling n. For small t', d_xy-wave correlations dominate across the entire filling range. In contrast, for larger t', the dominant pairing symmetry varies with n, reflecting a nontrivial interplay between frustration and density. These findings establish carrier filling and NNN hopping as complementary levers for manipulating CPBM stability and provide concrete criteria for identifying non-superfluid bosonic matter in cold-atom and correlated-electron systems.

physics.comp-ph↗

arXiv2024

FullStack Bench: Evaluating LLMs as Full Stack Coders

As the capabilities of code large language models (LLMs) continue to expand, their applications across diverse code intelligence domains are rapidly increasing. However, most existing datasets only evaluate limited application domains. To address this gap, we have developed a comprehensive code evaluation dataset FullStack Bench focusing on full-stack programming, which encompasses a wide range of application domains (e.g., basic programming, data analysis, software engineering, mathematics, and machine learning). Besides, to assess multilingual programming capabilities, in FullStack Bench, we design real-world instructions and corresponding unit test cases from 16 widely-used programming languages to reflect real-world usage scenarios rather than simple translations. Moreover, we also release an effective code sandbox execution tool (i.e., SandboxFusion) supporting various programming languages and packages to evaluate the performance of our FullStack Bench efficiently. Comprehensive experimental results on our FullStack Bench demonstrate the necessity and effectiveness of our FullStack Bench and SandboxFusion.

cs.AI↗

arXiv2024

Efficient sampling using Macrocanonical Monte Carlo and density of states mapping

In the context of Monte Carlo sampling for lattice models, the complexity of the energy landscape often leads to Markov chains being trapped in local optima, thereby increasing the correlation between samples and reducing sampling efficiency. This study proposes a Monte Carlo algorithm that effectively addresses the irregularities of the energy landscape through the introduction of the estimated density of states. This algorithm enhances the accuracy in the study of phase transitions and is not model-specific. Although our algorithm is primarily demonstrated on the two-dimensional square lattice model, the method is also applicable to a broader range of lattice and higher-dimensional models. Furthermore, the study develops a method for estimating the density of states of large systems based on that of smaller systems, enabling high-precision density of states estimation within specific energy intervals in large systems without sampling. For regions of lower precision, a re-weighting strategy is employed to adjust the density of states to enhance the precision further. This algorithm is not only significant within the field of lattice model sampling but may also inspire applications of the Monte Carlo method in other domains.

cond-mat.stat-mech↗

arXiv2024

Obtaining Optimal Spiking Neural Network in Sequence Learning via CRNN-SNN Conversion

Spiking neural networks (SNNs) are becoming a promising alternative to conventional artificial neural networks (ANNs) due to their rich neural dynamics and the implementation of energy-efficient neuromorphic chips. However, the non-differential binary communication mechanism makes SNN hard to converge to an ANN-level accuracy. When SNN encounters sequence learning, the situation becomes worse due to the difficulties in modeling long-range dependencies. To overcome these difficulties, researchers developed variants of LIF neurons and different surrogate gradients but still failed to obtain good results when the sequence became longer (e.g., $>$500). Unlike them, we obtain an optimal SNN in sequence learning by directly mapping parameters from a quantized CRNN. We design two sub-pipelines to support the end-to-end conversion of different structures in neural networks, which is called CNN-Morph (CNN $\rightarrow$ QCNN $\rightarrow$ BIFSNN) and RNN-Morph (RNN $\rightarrow$ QRNN $\rightarrow$ RBIFSNN). Using conversion pipelines and the s-analog encoding method, the conversion error of our framework is zero. Furthermore, we give the theoretical and experimental demonstration of the lossless CRNN-SNN conversion. Our results show the effectiveness of our method over short and long timescales tasks compared with the state-of-the-art learning- and conversion-based methods. We reach the highest accuracy of 99.16% (0.46 $\uparrow$) on S-MNIST, 94.95% (3.95 $\uparrow$) on PS-MNIST (sequence length of 784) respectively, and the lowest loss of 0.057 (0.013 $\downarrow$) within 8 time-steps in collision avoidance dataset.

cs.AI↗

arXiv2024

p-wave superconductivity induced by nearest-neighbor attraction in the square-lattice extended Hubbard model

The two-dimensional (2D) Hubbard model is widely believed to contain the key ingredients of high-temperature superconductivity in cuprate materials. Here, we report a constrained path quantum Monte Carlo (CPQMC) study of the square-lattice extended Hubbard model with on-site Coulomb repulsion U and nearest-neighbor (NN) electron attraction V. Upon doping $\delta$= 0.125, we find that the NN electron attraction V can notably drive an exotic spin-triplet (p-wave) superconducting (SC) phase, and enhance the p-wave SC correlations with the increase of V. But in the intermediate coupling regime, the $d_{x^2-y^2}$-wave (d-wave) does not significantly increase with the increase of V, indicating that the d-wave is not affected by V in strongly correlated system. Besides the pairing phase, a spin density wave (SDW) only exists near the half-filling in the particle-hole channel, and doping disrupts the formation of SDW order. Especially, the NN electron attraction V has no significant effect on SDW, reflecting the consistent relationship between d-wave SC and spin correlation. Moreover, as doping increases, the dominant region of p-wave also expands, further suppressing the presence of d-wave, which may help explain the disappearance of d-wave SC in overdoped cuprate superconductors. We also find the d-wave exhibits a singular nonzero (near point ($2{\pi}/3, {\pi}$)) condensation structure in momentum space, resulting in a different staggered behavior in the x and y direction with distance in real space. On the contrary, the p-wave condensed at zero momentum, and the p-wave correlation exhibits exponential decay in real space. Our work suggests the p-wave SC region can be induced and further broadened by the NN electron attraction V, offering a feasible mechanism to realize p-wave superconductivity in realistic cuprate materials.

cond-mat.supr-con↗

arXiv2024

dx2-y2-wave Bose Metal induced by the next-nearest-neighbor hopping t'

Superconductivity arises when electrons form Cooper pairs with phase coherence. In contrast, a lack of phase coherence in Cooper pairs can lead to an uncondensed metallic ground state known as the Bose metal state. In this study, we investigate an attractively interacting fermionic system with nearest-neighbor (NN) hopping (t) and next-nearest-neighbor (NNN) hopping (t') anisotropy between two species of spins in a two-dimensional (2D) lattice. Utilizing the constrained path quantum Monte Carlo (CPQMC) method, we demonstrate the existence of a dx2-y2-wave Cooper pair Bose metal (CPBM) phase with t'/t > 0.7. The CPBM phase exhibits a dome-like structure in the phase diagram of filling n~0.65, with the maximal region around an optimal t'/t ~ 0.2, suggesting that an appropriate value of t' facilitates the formation of the Bose metal. Furthermore, we find that a Bose metal formed by fermions with a closed Fermi surface confirms that the crucial condition for this exotic phenomenon is primarily the anisotropy of the Fermi surface, rather than its topology. Our finding of the dx2-y2-wave CPBM demonstrates the same pairing symmetry as the pseudogap behavior in cuprates, and its experimental realization in ultracold atom systems is also feasible.

cond-mat.str-el↗

arXiv2024

Exotic d-wave Cooper Pair Bose Metal in two dimensions

The study of non-Fermi liquids sheds light on unconventional phenomena in condensed matter systems that lie beyond the scope of Landau Fermi liquid theory. One intriguing example is the Bose metal, characterized by an uncondensed bosonic ground state. However, constructing a Bose metal phase in two dimensions (2D) remains a significant challenge.} Utilizing constraint path quantum Monte Carlo and functional renormalization group methods on a fermionic system with spin anisotropy in a 2D lattice, we reveal the emergence of a Cooper pair Bose metal (CPBM) phase in a highly anisotropic regime ($\alpha < 0.30$) with wide range of filling, as proposed in [A. E. Feiguin and M. P. A. Fisher, Phys. Rev. Lett. 103, 025303 (2009)]. Our findings exhibit a visible nonzero momentum Bose surface in the Cooper-pair distribution function, accompanied by a distinct signal of $d_{xy}$ correlation between pairs. Our results highlight that spin-dependent anisotropy in the Fermi surface leads to versatile pairing forms. Platforms such as ultracold atoms in optical lattices and recently proposed altermagnets hold promise for realizing this intriguing phase.

cond-mat.supr-con↗

arXiv2024

conv_einsum: A Framework for Representation and Fast Evaluation of Multilinear Operations in Convolutional Tensorial Neural Networks

Modern ConvNets continue to achieve state-of-the-art results over a vast array of vision and image classification tasks, but at the cost of increasing parameters. One strategy for compactifying a network without sacrificing much expressive power is to reshape it into a tensorial neural network (TNN), which is a higher-order tensorization of its layers, followed by a factorization, such as a CP-decomposition, which strips a weight down to its critical basis components. Passes through TNNs can be represented as sequences of multilinear operations (MLOs), where the evaluation path can greatly affect the number of floating point operations (FLOPs) incurred. While functions such as the popular einsum can evaluate simple MLOs such as contractions, existing implementations cannot process multi-way convolutions, resulting in scant assessments of how optimal evaluation paths through tensorized convolutional layers can improve training speed. In this paper, we develop a unifying framework for representing tensorial convolution layers as einsum-like strings and a meta-algorithm conv_einsum which is able to evaluate these strings in a FLOPs-minimizing manner. Comprehensive experiments, using our open-source implementation, over a wide range of models, tensor decompositions, and diverse tasks, demonstrate that conv_einsum significantly increases both computational and memory-efficiency of convolutional TNNs.

cs.LG↗

arXiv2023

Ab Initio Studies on Interactions in K$_3$C$_{60}$ under High Pressure

Fullerene solids doped with alkali metals (A$_3$C$_{60}$, A = K, Rb, Cs) exhibit a superconducting transition temperature ($T_c$) as high as 40 K, and their unconventional superconducting properties have been a subject of debate. With application of high pressure on K$_3$C$_{60}$ and Rb$_3$C$_{60}$, the experiments demonstrate the decrease of $T_c$. In this paper, we focus on K$_3$C$_{60}$ and derive the structure of K$_3$C$_{60}$ under different pressures based on first-principles calculations, exploring the trends of Coulomb interactions at various pressures. By utilizing the Maximally Localized Wannier function approach, Constrained Density Functional Perturbation Theory (cDFPT), and Constrained Random Phase Approximation (cRPA), we construct a microscopic low-energy model near the Fermi level. Our results strongly indicate that, in the K$_3$C$_{60}$ system, as pressure increases, the effect of phonons is the key to intraorbital electron pairing. There is a dominance of the phonon-driven superconducting mechanism at high pressure.

cond-mat.supr-con↗

arXiv2023

LEMON: Lossless model expansion

Scaling of deep neural networks, especially Transformers, is pivotal for their surging performance and has further led to the emergence of sophisticated reasoning capabilities in foundation models. Such scaling generally requires training large models from scratch with random initialization, failing to leverage the knowledge acquired by their smaller counterparts, which are already resource-intensive to obtain. To tackle this inefficiency, we present $\textbf{L}$ossl$\textbf{E}$ss $\textbf{MO}$del Expansio$\textbf{N}$ (LEMON), a recipe to initialize scaled models using the weights of their smaller but pre-trained counterparts. This is followed by model training with an optimized learning rate scheduler tailored explicitly for the scaled models, substantially reducing the training time compared to training from scratch. Notably, LEMON is versatile, ensuring compatibility with various network structures, including models like Vision Transformers and BERT. Our empirical results demonstrate that LEMON reduces computational costs by 56.7% for Vision Transformers and 33.2% for BERT when compared to training from scratch.

cs.LG↗

arXiv2023

Reviving Shift Equivariance in Vision Transformers

Shift equivariance is a fundamental principle that governs how we perceive the world - our recognition of an object remains invariant with respect to shifts. Transformers have gained immense popularity due to their effectiveness in both language and vision tasks. While the self-attention operator in vision transformers (ViT) is permutation-equivariant and thus shift-equivariant, patch embedding, positional encoding, and subsampled attention in ViT variants can disrupt this property, resulting in inconsistent predictions even under small shift perturbations. Although there is a growing trend in incorporating the inductive bias of convolutional neural networks (CNNs) into vision transformers, it does not fully address the issue. We propose an adaptive polyphase anchoring algorithm that can be seamlessly integrated into vision transformer models to ensure shift-equivariance in patch embedding and subsampled attention modules, such as window attention and global subsampled attention. Furthermore, we utilize depth-wise convolution to encode positional information. Our algorithms enable ViT, and its variants such as Twins to achieve 100% consistency with respect to input shift, demonstrate robustness to cropping, flipping, and affine transformations, and maintain consistent predictions even when the original models lose 20 percentage points on average when shifted by just a few pixels with Twins' accuracy dropping from 80.57% to 62.40%.

cs.CV↗

arXiv2022

Adversarial Auto-Augment with Label Preservation: A Representation Learning Principle Guided Approach

Data augmentation is a critical contributing factor to the success of deep learning but heavily relies on prior domain knowledge which is not always available. Recent works on automatic data augmentation learn a policy to form a sequence of augmentation operations, which are still pre-defined and restricted to limited options. In this paper, we show that a prior-free autonomous data augmentation's objective can be derived from a representation learning principle that aims to preserve the minimum sufficient information of the labels. Given an example, the objective aims at creating a distant "hard positive example" as the augmentation, while still preserving the original label. We then propose a practical surrogate to the objective that can be optimized efficiently and integrated seamlessly into existing methods for a broad class of machine learning tasks, e.g., supervised, semi-supervised, and noisy-label learning. Unlike previous works, our method does not require training an extra generative model but instead leverages the intermediate layer representations of the end-task model for generating data augmentations. In experiments, we show that our method consistently brings non-trivial improvements to the three aforementioned learning tasks from both efficiency and final performance, either or not combined with strong pre-defined augmentations, e.g., on medical images when domain knowledge is unavailable and the existing augmentation techniques perform poorly. Code is available at: https://github.com/kai-wen-yang/LPA3}{https://github.com/kai-wen-yang/LPA3.

cs.LG↗

arXiv2021

Certified Defense via Latent Space Randomized Smoothing with Orthogonal Encoders

Randomized Smoothing (RS), being one of few provable defenses, has been showing great effectiveness and scalability in terms of defending against $\ell_2$-norm adversarial perturbations. However, the cost of MC sampling needed in RS for evaluation is high and computationally expensive. To address this issue, we investigate the possibility of performing randomized smoothing and establishing the robust certification in the latent space of a network, so that the overall dimensionality of tensors involved in computation could be drastically reduced. To this end, we propose Latent Space Randomized Smoothing. Another important aspect is that we use orthogonal modules, whose Lipschitz property is known for free by design, to propagate the certified radius estimated in the latent space back to the input space, providing valid certifiable regions for the test samples in the input space. Experiments on CIFAR10 and ImageNet show that our method achieves competitive certified robustness but with a significant improvement of efficiency during the test phase.

cs.LG↗