arXiv subjects

Ling Tang

Publications and source records attributed to Ling Tang.

At least 19 recordsLinked to original sources

arXiv2026

Complex crystal structure prediction using ML-enhanced multi-minima iterative genetic algorithm

Current machine learning (ML) approaches for materials discovery rely heavily on known structural databases, limiting their ability to identify entirely novel structure types. In this work, we develop a multi-minima iterative genetic algorithm (MMIGA) that integrates an artificial-neural-network machine learning (ANN-ML) interatomic potential with an iterative, metadynamics-inspired penalty scheme. We demonstrate the robustness of this method on a complex ternary La-Co-Pb system, characterized by Co-Pb immiscibility and an intricate energy landscape. The ML-enhanced MMIGA successfully predicts the ground-state Pbam structure of the recently synthesized La4Co4Pb antagonistic-pair-phase, a novel structure missed by previous database-reliant ML predictions, while also identifying multiple metastable competing phases. Additionally, we challenged the MMIGA method to predict the structure of La5CoPb2 antagonistic-pair-phase, a new compound discovered during earlier attempts to synthesize the predicted phase La3CoPb. With only knowledge of the composition, our MMIGA approach successfully predicts the orthorhombic structure of La5CoPb2, producing an exact match with the structure independently determined by x-ray diffraction. By efficiently mapping both global minimum and relevant competing metastable states, this approach provides critical theoretical insights into phase selection for novel quantum and magnetic materials.

cond-mat.mtrl-sci↗

arXiv2026

AgentDoG 1.5: A Lightweight and Scalable Alignment Framework for AI Agent Safety and Security

Modern open-world agents such as OpenClaw exhibit powerful cross-environment execution capabilities yet introduce broad new safety risk sources. Meanwhile, advanced frontier AI models drastically lower attack barriers, rendering current agent alignment frameworks inadequate for real-world deployment. To tackle these emerging threats, we propose a lightweight and scalable agent safety alignment framework. Specifically, we update the agent safety taxonomy to accommodate emergent risks from Codex and OpenClaw execution scenarios. We further build a taxonomy-guided data engine with influence-function purification to train lightweight AgentDoG 1.5 variants (0.8B, 2B, 4B, and 8B parameters) using only around 1k samples, achieving comparable performance with leading closed-source models (e.g., GPT-5.4). Based on AgentDoG 1.5, we construct a highly efficient agentic safety SFT and RL training environment, which reduces deployment overhead in Docker-level environments by two orders of magnitude. Finally, we deploy AgentDoG 1.5 as a training-free online guardrail for real-time safety moderation. Extensive experimental results indicate that AgentDoG 1.5 achieves state-of-the-art performance in diverse and complex interactive agentic scenarios. All models and datasets are openly released.

cs.AI↗

arXiv2026

Attributing Emergence in Million-Agent Systems

Large language models (LLMs) can simulate human-like reasoning and decision-making in individual agents. LLM-powered multi-agent systems (MAS) combine such agents to simulate population-scale social phenomena such as polarization, information cascades, and market panics. Such studies require attributing macro emergence to individual agents, but existing axiomatic methods scale combinatorially in $N$ and have been confined to $N \lesssim 10^3$, while the phenomena they explain occur at $N \geq 10^6$. We address this gap by adapting Aumann--Shapley path-integral attribution to LLM-powered MAS at million-agent scale; the resulting method satisfies all four axioms, runs three to five orders of magnitude faster than sampled Shapley on the same hardware, and extends feasible axiomatic attribution by over three orders of magnitude (a $1670\times$ jump). We use this method to test the scale gap empirically: across 14 days of public Bluesky data ($1{,}671{,}587$ active users, five topics), we compute the attribution at both full scale and the visibility-biased $N = 10^2$ convenience sample used by small-scale studies, and the two disagree structurally. At full scale the long tail and middle tier jointly carry the majority; the biased small panel shifts about twice that share onto the upper follower tiers ($48\%$ versus $24\%$). We then prove that the disagreement cannot in general be reduced by post-hoc rescaling: an Attribution Scaling Bias theorem shows that a reconciling global rescaling factor exists exactly when the macro indicator is linear over agents, and our nonlinear indicators give residuals of $0.10$--$0.98$. For such nonlinear indicators, full-scale attribution is therefore a requirement rather than a methodological choice.

cs.AI↗

arXiv2026

What Do EEG Foundation Models Capture from Human Brain Signals?

Clinical electroencephalogram (EEG) analysis rests on a hand-crafted feature catalog refined over decades, \emph{e.g.,} band power, connectivity, complexity, and more. Modern EEG foundation models bypass this catalog, learn directly from raw signals via self-supervised pretraining, and match or outperform feature-engineered baselines on most clinical benchmarks. Whether the two representations align is an open question, which we decompose into three sub-questions: \emph{what does the model learn}, \emph{what does the model use}, and \emph{how much can be explained}. We answer them with layer-wise ridge probing, LEACE-style cross-covariance subspace erasure, and a transparent classifier benchmarked against a random-feature baseline. The audit covers three foundation models (CSBrain, CBraMod, LaBraM), five clinical tasks (MDD, Stress, ISRUC-Sleep, TUSL, Siena), and a 6-family 63-feature lexicon. Of the $945$ (model, task, feature) units, $648$ ($68.6\%$) are representation-causal and $199$ ($21.1\%$) are encoded-only. Across tasks, $50$ features qualify as universal candidates with strong support (all three architectures RC) in two or more tasks. Frequency-domain features dominate, but the other five families each contribute substantial causal mass. Confirmed features recover, on average, $79.3\%$ of the foundation model's advantage over the random baseline, with a clean task gradient (MDD $\approx 0.99$ down to Stress $\approx 0.56$): tasks near ceiling are almost fully recovered by the lexicon, while harder tasks leave a non-trivial residual that pinpoints a concrete target for future concept discovery.

cs.AI↗

arXiv2026

Interpreting Emergent Extreme Events in Multi-Agent Systems

Large language model-powered multi-agent systems have emerged as powerful tools for simulating complex human-like systems. The interactions within these systems often lead to extreme events whose origins remain obscured by the black box of emergence. Interpreting these events is critical for system safety. This paper proposes the first framework for explaining emergent extreme events in multi-agent systems, aiming to answer three fundamental questions: When does the event originate? Who drives it? And what behaviors contribute to it? Specifically, we adapt the Shapley value to faithfully attribute the occurrence of extreme events to each action taken by agents at different time steps, i.e., assigning an attribution score to the action to measure its influence on the event. We then aggregate the attribution scores along the dimensions of time, agent, and behavior to quantify the risk contribution of each dimension. Finally, we design a set of metrics based on these contribution scores to characterize the features of extreme events. Experiments across diverse multi-agent system scenarios (economic, financial, and social) demonstrate the effectiveness of our framework and provide general insights into the emergence of extreme phenomena.

cs.MA↗

arXiv2026

AgentDoG: A Diagnostic Guardrail Framework for AI Agent Safety and Security

The rise of AI agents introduces complex safety and security challenges arising from autonomous tool use and environmental interactions. Current guardrail models lack agentic risk awareness and transparency in risk diagnosis. To introduce an agentic guardrail that covers complex and numerous risky behaviors, we first propose a unified three-dimensional taxonomy that orthogonally categorizes agentic risks by their source (where), failure mode (how), and consequence (what). Guided by this structured and hierarchical taxonomy, we introduce a new fine-grained agentic safety benchmark (ATBench) and a Diagnostic Guardrail framework for agent safety and security (AgentDoG). AgentDoG provides fine-grained and contextual monitoring across agent trajectories. More Crucially, AgentDoG can diagnose the root causes of unsafe actions and seemingly safe but unreasonable actions, offering provenance and transparency beyond binary labels to facilitate effective agent alignment. AgentDoG variants are available in three sizes (4B, 7B, and 8B parameters) across Qwen and Llama model families. Extensive experimental results demonstrate that AgentDoG achieves state-of-the-art performance in agentic safety moderation in diverse and complex interactive scenarios. All models and datasets are openly released.

cs.AI↗

arXiv2026

The Why Behind the Action: Unveiling Internal Drivers via Agentic Attribution

Large Language Model (LLM)-based agents are widely used in real-world applications such as customer service, web navigation, and software engineering. As these systems become more autonomous and are deployed at scale, understanding why an agent takes a particular action becomes increasingly important for accountability and governance. However, existing research predominantly focuses on \textit{failure attribution} to localize explicit errors in unsuccessful trajectories, which is insufficient for explaining \textbf{the reason behind agent behaviors}. To bridge this gap, we propose a novel framework for \textbf{general agentic attribution}, designed to identify the internal factors driving agent actions regardless of the task outcome. Our framework operates hierarchically to manage the complexity of agent interactions. Specifically, at the \textit{component level}, we employ temporal likelihood dynamics to identify critical interaction steps; then at the \textit{sentence level}, we refine this localization using perturbation-based analysis to isolate the specific textual evidence. We validate our framework across a diverse suite of agentic scenarios, including standard tool use and subtle reliability risks like memory-induced bias. Experimental results demonstrate that the proposed framework reliably pinpoints pivotal historical events and sentences behind the agent behavior, offering a critical step toward safer and more accountable agentic systems. Codes are available at https://github.com/AI45Lab/AgentDoG.

cs.AI↗

arXiv2025

Developing a Neural Network Machine Learning Interatomic Potential for Molecular Dynamics Simulations of La-Si-P Systems

While molecular dynamics (MD) is a very useful computational method for atomistic simulations, modeling the interatomic interactions for reliable MD simulations of real materials has been a long-standing challenge. In 2007, Behler and Perrinello first proposed and demonstrated an artificial neural network machine learning (ANN-ML) scheme, opening a new paradigm for developing accurate and efficient interatomic potentials for reliable MD simulation studies of the thermodynamics and kinetics of materials. In this paper, we show that an accurate and transferable ANN-ML interatomic potential can be developed for MD simulations of La-Si-P system. The crucial role of training data in the ML potential development is discussed. The developed ANN-ML potential accurately describes not only the energy versus volume curves for all the known elemental, binary, and ternary crystalline structures in La-Si-P system, but also the structures of La-Si-P liquids with various compositions. Using the developed ANN-ML potential, the melting temperatures of several crystalline phases in La-Si-P system are predicted by the coexistence of solid-liquid phases from MD simulations. While the ANN-ML model systematically underestimates the melting temperatures of these phases, the overall trend agrees with experiment. The developed ANN-ML potential is also applied to study the nucleation and growth of LaP as a function of different relative concentrations of Si and P in the La-Si-P liquid, and the obtained results are consistent with experimental observations.

cond-mat.mtrl-sci↗

arXiv2025

DecIF: Improving Instruction-Following through Meta-Decomposition

Instruction-following has emerged as a crucial capability for large language models (LLMs). However, existing approaches often rely on pre-existing documents or external resources to synthesize instruction-following data, which limits their flexibility and generalizability. In this paper, we introduce DecIF, a fully autonomous, meta-decomposition guided framework that generates diverse and high-quality instruction-following data using only LLMs. DecIF is grounded in the principle of decomposition. For instruction generation, we guide LLMs to iteratively produce various types of meta-information, which are then combined with response constraints to form well-structured and semantically rich instructions. We further utilize LLMs to detect and resolve potential inconsistencies within the generated instructions. Regarding response generation, we decompose each instruction into atomic-level evaluation criteria, enabling rigorous validation and the elimination of inaccurate instruction-response pairs. Extensive experiments across a wide range of scenarios and settings demonstrate DecIF's superior performance on instruction-following tasks. Further analysis highlights its strong flexibility, scalability, and generalizability in automatically synthesizing high-quality instruction data.

cs.CL↗

arXiv2025

Towards the Resistance of Neural Network Watermarking to Fine-tuning

This paper proves a new watermarking method to embed the ownership information into a deep neural network (DNN), which is robust to fine-tuning. Specifically, we prove that when the input feature of a convolutional layer only contains low-frequency components, specific frequency components of the convolutional filter will not be changed by gradient descent during the fine-tuning process, where we propose a revised Fourier transform to extract frequency components from the convolutional filter. Additionally, we also prove that these frequency components are equivariant to weight scaling and weight permutations. In this way, we design a watermark module to encode the watermark information to specific frequency components in a convolutional filter. Preliminary experiments demonstrate the effectiveness of our method.

cs.LG↗

arXiv2025

CIMFlow: An Integrated Framework for Systematic Design and Evaluation of Digital CIM Architectures

Digital Compute-in-Memory (CIM) architectures have shown great promise in Deep Neural Network (DNN) acceleration by effectively addressing the "memory wall" bottleneck. However, the development and optimization of digital CIM accelerators are hindered by the lack of comprehensive tools that encompass both software and hardware design spaces. Moreover, existing design and evaluation frameworks often lack support for the capacity constraints inherent in digital CIM architectures. In this paper, we present CIMFlow, an integrated framework that provides an out-of-the-box workflow for implementing and evaluating DNN workloads on digital CIM architectures. CIMFlow bridges the compilation and simulation infrastructures with a flexible instruction set architecture (ISA) design, and addresses the constraints of digital CIM through advanced partitioning and parallelism strategies in the compilation flow. Our evaluation demonstrates that CIMFlow enables systematic prototyping and optimization of digital CIM architectures across diverse configurations, providing researchers and designers with an accessible platform for extensive design space exploration.

cs.AR↗

arXiv2023

Structure and dynamics of Fe90Si3O7 liquids close to Earth's liquid core conditions

Using an artificial neural-network machine learning interatomic potential, we have performed molecular dynamics simulations to study the structure and dynamics of Fe90Si3O7 liquid close to the Earth's liquid core conditions. The simulation results reveal that the short-range structural order (SRO) in the Fe90Si3O7 liquid is very strong. About 80% of the atoms are arranged in crystalline-like SRO motifs. In particular, ~70% of Fe-centered clusters can be classified as either hexagonal-close-pack (HCP/HCP-like) or icosahedral (ICO/ICO-like) SRO motifs. The SRO clusters centered on Fe, Si, or O atoms are strongly intermixed and homogenously distributed throughout the liquid. The atomic structure of the liquid and the fractions of dominant SRO clusters are not sensitive to pressure/temperature used in the simulations except that the SRO of the O-centered clusters is enhanced close to inner core pressures. The O diffusion coefficient is about 2-3 times larger than the Fe and Si ions and increases more rapidly in the deeper core regions.

cond-mat.mtrl-sci↗

arXiv2022

Subsampling for tensor least squares: Optimization and statistical perspectives

In this paper, we investigate the random subsampling method for tensor least squares problem with respect to the popular t-product. From the optimization perspective, we present the error bounds in the sense of probability for the residual and solution obtained by the proposed method. From the statistical perspective, we derive the expressions of the conditional and unconditional expectations and variances for the solution, where the unconditional ones combine the model noises. Moreover, based on the unconditional variance, an optimal subsampling probability distribution is also found. Finally, the feasibility and effectiveness of the proposed method and the correctness of the theoretical results are verified by numerical experiments.

math.NA↗

arXiv2022

Defects of Convolutional Decoder Networks in Frequency Representation

In this paper, we prove the representation defects of a cascaded convolutional decoder network, considering the capacity of representing different frequency components of an input sample. We conduct the discrete Fourier transform on each channel of the feature map in an intermediate layer of the decoder network. Then, we extend the 2D circular convolution theorem to represent the forward and backward propagations through convolutional layers in the frequency domain. Based on this, we prove three defects in representing feature spectrums. First, we prove that the convolution operation, the zero-padding operation, and a set of other settings all make a convolutional decoder network more likely to weaken high-frequency components. Second, we prove that the upsampling operation generates a feature spectrum, in which strong signals repetitively appear at certain frequencies. Third, we prove that if the frequency components in the input sample and frequency components in the target output for regression have a small shift, then the decoder usually cannot be effectively learned.

cs.LG↗

arXiv2022

On sketch-and-project methods for solving tensor equations

We first propose the regular sketch-and-project method for solving tensor equations with respect to the popular t-product. Then, three adaptive sampling strategies and three corresponding adaptive sketch-and-project methods are derived. We prove that all the proposed methods have linear convergence in expectation. Furthermore, we investigate the Fourier domain versions and some special cases of the new methods, where the latter corresponds to some famous matrix equation methods. Finally, numerical experiments are presented to demonstrate and test the feasibility and effectiveness of the proposed methods for solving tensor equations

math.NA↗

arXiv2022

Greedy randomized sampling nonlinear Kaczmarz methods

The nonlinear Kaczmarz method was recently proposed to solve the system of nonlinear equations. In this paper, we first discuss two greedy selection rules, i.e., the maximum residual and maximum distance rules, for the nonlinear Kaczmarz iteration. Then, based on them, two kinds of greedy randomized sampling methods are presented. Further, we also devise four corresponding greedy randomized block methods, i.e., the multiple samples-based methods. The linear convergence in expectation of all the proposed methods is proved. Numerical results show that, in some applications including brown almost linear function and generalized linear model, the greedy selection rules give faster convergence rates than the random ones, and the block methods outperform the single sample-based ones.

math.NA↗

arXiv2022

Batch Normalization Is Blind to the First and Second Derivatives of the Loss

In this paper, we prove the effects of the BN operation on the back-propagation of the first and second derivatives of the loss. When we do the Taylor series expansion of the loss function, we prove that the BN operation will block the influence of the first-order term and most influence of the second-order term of the loss. We also find that such a problem is caused by the standardization phase of the BN operation. Experimental results have verified our theoretical conclusions, and we have found that the BN operation significantly affects feature representations in specific tasks, where losses of different samples share similar analytic formulas.

cs.LG↗

arXiv2022

A deep machine learning potential for atomistic simulation of Fe-Si-O systems under Earth's outer core conditions

Using artificial neural-network machine learning (ANN-ML) to generate interatomic potentials has been demonstrated to be a promising approach to address the long-standing challenge of accuracy versus efficiency in molecular dynamics (MD) simulations. Here, taking the Fe-Si-O system as a prototype, we show that accurate and transferable ANN-ML potentials can be developed for reliable MD simulations of materials at high-pressure and high-temperature conditions of the Earth's outer core. The ANN-ML potential for Fe-Si-O system is trained by fitting to the energies and forces of related binaries and ternary liquid structures at high pressures and temperatures obtained by first-principles calculations based on density functional theory (DFT). We show that the generated ANN-ML potential describes well the structure and dynamics of liquid phases of this complex system. The efficient ANN-ML potential with DFT accuracy provides a promising scheme for accurate atomistic simulations of structures and dynamics of complex Fe-Si-O system in the Earth's outer core.

cond-mat.mtrl-sci↗