arXiv subjects

Yifei Yang

Publications and source records attributed to Yifei Yang.

At least 55 records · Page 3Linked to original sources

arXiv2025

LESA: Learnable LLM Layer Scaling-Up

Training Large Language Models (LLMs) from scratch requires immense computational resources, making it prohibitively expensive. Model scaling-up offers a promising solution by leveraging the parameters of smaller models to create larger ones. However, existing depth scaling-up methods rely on empirical heuristic rules for layer duplication, which result in poorer initialization and slower convergence during continual pre-training. We propose \textbf{LESA}, a novel learnable method for depth scaling-up. By concatenating parameters from each layer and applying Singular Value Decomposition, we uncover latent patterns between layers, suggesting that inter-layer parameters can be learned. LESA uses a neural network to predict the parameters inserted between adjacent layers, enabling better initialization and faster training. Experiments show that LESA outperforms existing baselines, achieving superior performance with less than half the computational cost during continual pre-training. Extensive analyses demonstrate its effectiveness across different model sizes and tasks.

cs.LG↗

arXiv2025

Enhancing Unconventional Spin-Orbit Torque Efficiency: Numerical Study on the Influence of Crystallographic Texture and Polycrystalline Effects on Low-Symmetry Materials

Spin-orbit torque (SOT) has been extensively studied as a key mechanism in spintronics applications. However, conventional SOT materials limit the spin polarization direction to the in-plane orientation, which is suboptimal for efficient magnetization switching. Recently, spin currents with spin polarization along multiple directions have been observed in low-symmetry materials, offering a promising energy-efficient strategy for the field-free switching of magnetic materials with perpendicular magnetic anisotropy. However, the efficiency of this mechanism is highly dependent on the crystallographic texture of the SOT materials, a critical factor that, to date, has not been quantitatively investigated. In this study, we present a comprehensive numerical investigation into the impact of both in-plane and out-of-plane crystallographic textures of SOT materials on the unconventional SOT generated by Dresselhaus-like and out-of-plane spin polarizations. By employing a theoretical orientation distribution function, we calculate the effective unconventional SOT values for SOT materials with tunable crystallographic texture. This analysis provides a framework for the synthesis and optimization of future low-symmetry SOT materials, which can enhance operational efficiency for spintronics applications in magnetoresistive random-access memory and spin logic devices.

cond-mat.mes-hall↗

arXiv2024

SCANS: Mitigating the Exaggerated Safety for LLMs via Safety-Conscious Activation Steering

Safety alignment is indispensable for Large Language Models (LLMs) to defend threats from malicious instructions. However, recent researches reveal safety-aligned LLMs prone to reject benign queries due to the exaggerated safety issue, limiting their helpfulness. In this paper, we propose a Safety-Conscious Activation Steering (SCANS) method to mitigate the exaggerated safety concerns in aligned LLMs. First, SCANS extracts the refusal steering vectors within the activation space and utilizes vocabulary projection to anchor some specific safety-critical layers which influence model refusal behavior. Second, by tracking the hidden state transition, SCANS identifies the steering direction and steers the model behavior accordingly, achieving a balance between exaggerated safety and adequate safety. Experiments show that SCANS achieves new state-of-the-art performance on XSTest and OKTest benchmarks, without impairing their defense capability against harmful queries and maintaining almost unchanged model capability.

cs.AI↗

arXiv2024

Energy Efficient Stochastic Signal Manipulation in Superparamagnetic Tunnel Junctions via Voltage-Controlled Exchange Coupling

Superparamagnetic tunnel junctions (sMTJs) are emerging as promising components for stochastic units in neuromorphic computing, owing to their tunable random switching behavior. Conventional MTJ control methods, such as spin-transfer torque (STT) and spin-orbit torque (SOT), often require substantial power. Here, we introduce the voltage-controlled exchange coupling (VCEC) mechanism, enabling switching between antiparallel and parallel states in sMTJs with an ultralow power consumption of only 40 nW, approximately two orders of magnitude lower than conventional STT-based sMTJs. This mechanism yields a sigmoid-shaped output response, making it ideally suited for neuromorphic computing applications. Furthermore, we validate the feasibility of integrating VCEC with the SOT current control, offering an additional dimension for magnetic state manipulation. This work marks the first practical demonstration of VCEC effect in sMTJs, highlighting its potential as a low-power control solution for probabilistic bits in advanced computing systems.

physics.app-ph↗

arXiv2024

Giant spin Hall effect with multi-directional spin components in Ni4W

Spin-orbit torque (SOT) can be used to efficiently manipulate the magnetic state of magnetic materials, which is an essential element for memory and logic applications. Due to symmetry constraints, only in-plane spins can be injected into the ferromagnet from the underlying SOT layer for conventional SOT materials such as heavy metals and topological materials. Through the use of materials with low symmetries, or other symmetry breaking approaches, unconventional spin currents with out-of-plane polarization has been demonstrated and enabled field-free deterministic switching of perpendicular magnetization. Despite this progress, the SOT efficiency of these materials has typically remained low. Here, we report a giant SOT efficiency of 0.85 in sputtered Ni4W/CoFeB heterostructure at room temperature, as evaluated by second harmonic Hall measurements. In addition, due to the low crystal symmetry of Ni4W, unconventional out-of-plane and Dresselhaus-like spin components were observed. Macro-spin simulation suggests our spin Hall tensor to provide about an order of magnitude improvement in the magnetization switching efficiency, thus broadening the path towards energy efficient spintronic devices using low-symmetry materials.

cond-mat.mes-hall↗

arXiv2024

Head-wise Shareable Attention for Large Language Models

Large Language Models (LLMs) suffer from huge number of parameters, which restricts their deployment on edge devices. Weight sharing is one promising solution that encourages weight reuse, effectively reducing memory usage with less performance drop. However, current weight sharing techniques primarily focus on small-scale models like BERT and employ coarse-grained sharing rules, e.g., layer-wise. This becomes limiting given the prevalence of LLMs and sharing an entire layer or block obviously diminishes the flexibility of weight sharing. In this paper, we present a perspective on head-wise shareable attention for large language models. We further propose two memory-efficient methods that share parameters across attention heads, with a specific focus on LLMs. Both of them use the same dynamic strategy to select the shared weight matrices. The first method directly reuses the pre-trained weights without retraining, denoted as $\textbf{DirectShare}$. The second method first post-trains with constraint on weight matrix similarity and then shares, denoted as $\textbf{PostShare}$. Experimental results reveal our head-wise shared models still maintain satisfactory capabilities, demonstrating the feasibility of fine-grained weight sharing applied to LLMs.

cs.CL↗

arXiv2024

KVSharer: Efficient Inference via Layer-Wise Dissimilar KV Cache Sharing

The development of large language models (LLMs) has significantly expanded model sizes, resulting in substantial GPU memory requirements during inference. The key and value storage of the attention map in the KV (key-value) cache accounts for more than 80\% of this memory consumption. Nowadays, most existing KV cache compression methods focus on intra-layer compression within a single Transformer layer but few works consider layer-wise compression. In this paper, we propose a plug-and-play method called \textit{KVSharer}, which shares the KV cache between layers to achieve layer-wise compression. Rather than intuitively sharing based on higher similarity, we discover a counterintuitive phenomenon: sharing dissimilar KV caches better preserves the model performance. Experiments show that \textit{KVSharer} can reduce KV cache computation by 30\%, thereby lowering memory consumption without significantly impacting model performance and it can also achieve at least 1.3 times generation acceleration. Additionally, we verify that \textit{KVSharer} is compatible with existing intra-layer KV cache compression methods, and combining both can further save memory.

cs.LG↗

arXiv2024

LaCo: Large Language Model Pruning via Layer Collapse

Large language models (LLMs) based on transformer are witnessing a notable trend of size expansion, which brings considerable costs to both model training and inference. However, existing methods such as model quantization, knowledge distillation, and model pruning are constrained by various issues, including hardware support limitations, the need for extensive training, and alterations to the model internal structure. In this paper, we propose a concise layer-wise structured pruner called \textit{Layer Collapse (LaCo)}, in which rear model layers collapse into a prior layer, enabling a rapid reduction in model size while preserving the model structure. Comprehensive experiments show that our method maintains an average task performance of over 80\% at pruning ratios of 25-30\%, significantly outperforming existing state-of-the-art structured pruning methods. We also conduct post-training experiments to confirm that the \textit{LaCo} effectively inherits the parameters of the original model. Additionally, we perform ablation studies on various settings of \textit{LaCo}. Finally, we discuss our motivation from the perspective of layer-wise similarity and evaluate the performance of the pruned LLMs across various pruning ratios\footnote{\url{https://github.com/yangyifei729/LaCo}}.

cs.CL↗

arXiv2024

DORec: Decomposed Object Reconstruction and Segmentation Utilizing 2D Self-Supervised Features

Recovering 3D geometry and textures of individual objects is crucial for many robotics applications, such as manipulation, pose estimation, and autonomous driving. However, decomposing a target object from a complex background is challenging. Most existing approaches rely on costly manual labels to acquire object instance perception. Recent advancements in 2D self-supervised learning offer new prospects for identifying objects of interest, yet leveraging such noisy 2D features for clean decomposition remains difficult. In this paper, we propose a Decomposed Object Reconstruction (DORec) network based on neural implicit representations. Our key idea is to use 2D self-supervised features to create two levels of masks for supervision: a binary mask for foreground regions and a K-cluster mask for semantically similar regions. These complementary masks result in robust decomposition. Experimental results on different datasets show DORec's superiority in segmenting and reconstructing diverse foreground objects from varied backgrounds enabling downstream tasks such as pose estimation.

cs.CV↗

arXiv2024

BatGPT-Chem: A Foundation Large Model For Retrosynthesis Prediction

Retrosynthesis analysis is pivotal yet challenging in drug discovery and organic chemistry. Despite the proliferation of computational tools over the past decade, AI-based systems often fall short in generalizing across diverse reaction types and exploring alternative synthetic pathways. This paper presents BatGPT-Chem, a large language model with 15 billion parameters, tailored for enhanced retrosynthesis prediction. Integrating chemical tasks via a unified framework of natural language and SMILES notation, this approach synthesizes extensive instructional data from an expansive chemical database. Employing both autoregressive and bidirectional training techniques across over one hundred million instances, BatGPT-Chem captures a broad spectrum of chemical knowledge, enabling precise prediction of reaction conditions and exhibiting strong zero-shot capabilities. Superior to existing AI methods, our model demonstrates significant advancements in generating effective strategies for complex molecules, as validated by stringent benchmark tests. BatGPT-Chem not only boosts the efficiency and creativity of retrosynthetic analysis but also establishes a new standard for computational tools in synthetic design. This development empowers chemists to adeptly address the synthesis of novel compounds, potentially expediting the innovation cycle in drug manufacturing and materials science. We release our trial platform at \url{https://www.batgpt.net/dapp/chem}.

cs.LG↗

arXiv2024

Improving Hyperbolic Representations via Gromov-Wasserstein Regularization

Hyperbolic representations have shown remarkable efficacy in modeling inherent hierarchies and complexities within data structures. Hyperbolic neural networks have been commonly applied for learning such representations from data, but they often fall short in preserving the geometric structures of the original feature spaces. In response to this challenge, our work applies the Gromov-Wasserstein (GW) distance as a novel regularization mechanism within hyperbolic neural networks. The GW distance quantifies how well the original data structure is maintained after embedding the data in a hyperbolic space. Specifically, we explicitly treat the layers of the hyperbolic neural networks as a transport map and calculate the GW distance accordingly. We validate that the GW distance computed based on a training set well approximates the GW distance of the underlying data distribution. Our approach demonstrates consistent enhancements over current state-of-the-art methods across various tasks, including few-shot image classification, as well as semi-supervised graph link prediction and node classification.

cs.LG↗

arXiv2024

Monotone Generative Modeling via a Gromov-Monge Embedding

Generative adversarial networks (GANs) are popular for generative tasks; however, they often require careful architecture selection, extensive empirical tuning, and are prone to mode collapse. To overcome these challenges, we propose a novel model that identifies the low-dimensional structure of the underlying data distribution, maps it into a low-dimensional latent space while preserving the underlying geometry, and then optimally transports a reference measure to the embedded distribution. We prove three key properties of our method: 1) The encoder preserves the geometry of the underlying data; 2) The generator is $c$-cyclically monotone, where $c$ is an intrinsic embedding cost employed by the encoder; and 3) The discriminator's modulus of continuity improves with the geometric preservation of the data. Numerical experiments demonstrate the effectiveness of our approach in generating high-quality images and exhibiting robustness to both mode collapse and training instability.

cs.LG↗

arXiv2024

GRAM: An Interpretable Approach for Graph Anomaly Detection using Gradient Attention Maps

Detecting unusual patterns in graph data is a crucial task in data mining. However, existing methods face challenges in consistently achieving satisfactory performance and often lack interpretability, which hinders our understanding of anomaly detection decisions. In this paper, we propose a novel approach to graph anomaly detection that leverages the power of interpretability to enhance performance. Specifically, our method extracts an attention map derived from gradients of graph neural networks, which serves as a basis for scoring anomalies. Notably, our approach is flexible and can be used in various anomaly detection settings. In addition, we conduct theoretical analysis using synthetic data to validate our method and gain insights into its decision-making process. To demonstrate the effectiveness of our method, we extensively evaluate our approach against state-of-the-art graph anomaly detection techniques on real-world graph classification and wireless network datasets. The results consistently demonstrate the superior performance of our method compared to the baselines.

cs.LG↗

arXiv2024

Sparse Matrix Coding for URLLC

Sparse Vector Coding (SVC) has long been considered an encoding method that meets the URLLC QOS requirements. This encoding method has been widely studied and applied due to its low encoding and decoding complexity, no pilot transmission, resistance to inter-carrier interference, and low power consumption. However, due to the use of position indexing, the encoding essentially reduces the signal-to-noise ratio requirements by increasing the communication bandwidth, which also leads to low encoding efficiency and strong rigidity in decoding. Based on the sparse representation characteristics of SVC, we propose a joint sparse representation encoding, namely Sparse Matrix Coding (SMC). This encoding method utilizes multi-user information joint encoding, and the sparsity and sparse locations between users are shared.

eess.SP↗

arXiv2024

$ν$-DBA: Neural Implicit Dense Bundle Adjustment Enables Image-Only Driving Scene Reconstruction

The joint optimization of the sensor trajectory and 3D map is a crucial characteristic of bundle adjustment (BA), essential for autonomous driving. This paper presents $ν$-DBA, a novel framework implementing geometric dense bundle adjustment (DBA) using 3D neural implicit surfaces for map parametrization, which optimizes both the map surface and trajectory poses using geometric error guided by dense optical flow prediction. Additionally, we fine-tune the optical flow model with per-scene self-supervision to further improve the quality of the dense mapping. Our experimental results on multiple driving scene datasets demonstrate that our method achieves superior trajectory optimization and dense reconstruction accuracy. We also investigate the influences of photometric error and different neural geometric priors on the performance of surface reconstruction and novel view synthesis. Our method stands as a significant step towards leveraging neural implicit representations in dense bundle adjustment for more accurate trajectories and detailed environmental mapping.

cs.CV↗

arXiv2024

Drag Your Noise: Interactive Point-based Editing via Diffusion Semantic Propagation

Point-based interactive editing serves as an essential tool to complement the controllability of existing generative models. A concurrent work, DragDiffusion, updates the diffusion latent map in response to user inputs, causing global latent map alterations. This results in imprecise preservation of the original content and unsuccessful editing due to gradient vanishing. In contrast, we present DragNoise, offering robust and accelerated editing without retracing the latent map. The core rationale of DragNoise lies in utilizing the predicted noise output of each U-Net as a semantic editor. This approach is grounded in two critical observations: firstly, the bottleneck features of U-Net inherently possess semantically rich features ideal for interactive editing; secondly, high-level semantics, established early in the denoising process, show minimal variation in subsequent stages. Leveraging these insights, DragNoise edits diffusion semantics in a single denoising step and efficiently propagates these changes, ensuring stability and efficiency in diffusion editing. Comparative experiments reveal that DragNoise achieves superior control and semantic retention, reducing the optimization time by over 50% compared to DragDiffusion. Our codes are available at https://github.com/haofengl/DragNoise.

cs.CV↗

arXiv2024

Hypertext Entity Extraction in Webpage

Webpage entity extraction is a fundamental natural language processing task in both research and applications. Nowadays, the majority of webpage entity extraction models are trained on structured datasets which strive to retain textual content and its structure information. However, existing datasets all overlook the rich hypertext features (e.g., font color, font size) which show their effectiveness in previous works. To this end, we first collect a \textbf{H}ypertext \textbf{E}ntity \textbf{E}xtraction \textbf{D}ataset (\textit{HEED}) from the e-commerce domains, scraping both the text and the corresponding explicit hypertext features with high-quality manual entity annotations. Furthermore, we present the \textbf{Mo}E-based \textbf{E}ntity \textbf{E}xtraction \textbf{F}ramework (\textit{MoEEF}), which efficiently integrates multiple features to enhance model performance by Mixture of Experts and outperforms strong baselines, including the state-of-the-art small-scale models and GPT-3.5-turbo. Moreover, the effectiveness of hypertext features in \textit{HEED} and several model components in \textit{MoEEF} are analyzed.

cs.CL↗

arXiv2024

CMMLU: Measuring massive multitask language understanding in Chinese

As the capabilities of large language models (LLMs) continue to advance, evaluating their performance becomes increasingly crucial and challenging. This paper aims to bridge this gap by introducing CMMLU, a comprehensive Chinese benchmark that covers various subjects, including natural science, social sciences, engineering, and humanities. We conduct a thorough evaluation of 18 advanced multilingual- and Chinese-oriented LLMs, assessing their performance across different subjects and settings. The results reveal that most existing LLMs struggle to achieve an average accuracy of 50%, even when provided with in-context examples and chain-of-thought prompts, whereas the random baseline stands at 25%. This highlights significant room for improvement in LLMs. Additionally, we conduct extensive experiments to identify factors impacting the models' performance and propose directions for enhancing LLMs. CMMLU fills the gap in evaluating the knowledge and reasoning capabilities of large language models within the Chinese context.

cs.CL↗