arXiv subjects

Shuo Yang

Publications and source records attributed to Shuo Yang.

At least 271 records · Page 15Linked to original sources

arXiv2024

Large-scale Dataset Pruning with Dynamic Uncertainty

The state of the art of many learning tasks, e.g., image classification, is advanced by collecting larger datasets and then training larger models on them. As the outcome, the increasing computational cost is becoming unaffordable. In this paper, we investigate how to prune the large-scale datasets, and thus produce an informative subset for training sophisticated deep models with negligible performance drop. We propose a simple yet effective dataset pruning method by exploring both the prediction uncertainty and training dynamics. We study dataset pruning by measuring the variation of predictions during the whole training process on large-scale datasets, i.e., ImageNet-1K and ImageNet-21K, and advanced models, i.e., Swin Transformer and ConvNeXt. Extensive experimental results indicate that our method outperforms the state of the art and achieves 25% lossless pruning ratio on both ImageNet-1K and ImageNet-21K. The code and pruned datasets are available at https://github.com/BAAI-DCAI/Dataset-Pruning.

cs.LG↗

arXiv2024

S-LoRA: Serving Thousands of Concurrent LoRA Adapters

The "pretrain-then-finetune" paradigm is commonly adopted in the deployment of large language models. Low-Rank Adaptation (LoRA), a parameter-efficient fine-tuning method, is often employed to adapt a base model to a multitude of tasks, resulting in a substantial collection of LoRA adapters derived from one base model. We observe that this paradigm presents significant opportunities for batched inference during serving. To capitalize on these opportunities, we present S-LoRA, a system designed for the scalable serving of many LoRA adapters. S-LoRA stores all adapters in the main memory and fetches the adapters used by the currently running queries to the GPU memory. To efficiently use the GPU memory and reduce fragmentation, S-LoRA proposes Unified Paging. Unified Paging uses a unified memory pool to manage dynamic adapter weights with different ranks and KV cache tensors with varying sequence lengths. Additionally, S-LoRA employs a novel tensor parallelism strategy and highly optimized custom CUDA kernels for heterogeneous batching of LoRA computation. Collectively, these features enable S-LoRA to serve thousands of LoRA adapters on a single GPU or across multiple GPUs with a small overhead. Compared to state-of-the-art libraries such as HuggingFace PEFT and vLLM (with naive support of LoRA serving), S-LoRA can improve the throughput by up to 4 times and increase the number of served adapters by several orders of magnitude. As a result, S-LoRA enables scalable serving of many task-specific fine-tuned models and offers the potential for large-scale customized fine-tuning services. The code is available at https://github.com/S-LoRA/S-LoRA

cs.LG↗

arXiv2024

Data-efficient Fine-tuning for LLM-based Recommendation

Leveraging Large Language Models (LLMs) for recommendation has recently garnered considerable attention, where fine-tuning plays a key role in LLMs' adaptation. However, the cost of fine-tuning LLMs on rapidly expanding recommendation data limits their practical application. To address this challenge, few-shot fine-tuning offers a promising approach to quickly adapt LLMs to new recommendation data. We propose the task of data pruning for efficient LLM-based recommendation, aimed at identifying representative samples tailored for LLMs' few-shot fine-tuning. While coreset selection is closely related to the proposed task, existing coreset selection methods often rely on suboptimal heuristic metrics or entail costly optimization on large-scale recommendation data. To tackle these issues, we introduce two objectives for the data pruning task in the context of LLM-based recommendation: 1) high accuracy aims to identify the influential samples that can lead to high overall performance; and 2) high efficiency underlines the low costs of the data pruning process. To pursue the two objectives, we propose a novel data pruning method based on two scores, i.e., influence score and effort score, to efficiently identify the influential samples. Particularly, the influence score is introduced to accurately estimate the influence of sample removal on the overall performance. To achieve low costs of the data pruning process, we use a small-sized surrogate model to replace LLMs to obtain the influence score. Considering the potential gap between the surrogate model and LLMs, we further propose an effort score to prioritize some hard samples specifically for LLMs. Empirical results on three real-world datasets validate the effectiveness of our proposed method. In particular, the proposed method uses only 2% samples to surpass the full data fine-tuning, reducing time costs by 97%.

cs.IR↗

arXiv2024

Ultra-sensitive solid-state organic molecular microwave quantum receiver

High-accuracy microwave sensing is widely demanded in various fields, ranging from cosmology to microwave quantum technology. Quantum receivers based on inorganic solid-state spin systems are promising candidates for such purpose because of the stability and compatibility, but their best sensitivity is currently limited to a few pT/$\sqrt{\rm{Hz}}$. Here, by utilising an enhanced readout scheme with the state-of-the-art solid-state maser technology, we develop a robust microwave quantum receiver functioned by organic molecular spins at ambient conditions. Owing to the maser amplification, the sensitivity of the receiver achieves 6.14 $\pm$ 0.17 fT/$\sqrt{\rm{Hz}}$ which exceeds three orders of magnitude than that of the inorganic solid-state quantum receivers. The heterodyne detection without additional local oscillators improves bandwidth of the receiver and allows frequency detection. The scheme can be extended to other solid-state spin systems without complicated control pulses and thus enables practical applications such as electron spin resonance spectroscopy, dark matter searches, and astronomical observations.

quant-ph↗

arXiv2024

Revisiting Context Aggregation for Image Matting

Traditional studies emphasize the significance of context information in improving matting performance. Consequently, deep learning-based matting methods delve into designing pooling or affinity-based context aggregation modules to achieve superior results. However, these modules cannot well handle the context scale shift caused by the difference in image size during training and inference, resulting in matting performance degradation. In this paper, we revisit the context aggregation mechanisms of matting networks and find that a basic encoder-decoder network without any context aggregation modules can actually learn more universal context aggregation, thereby achieving higher matting performance compared to existing methods. Building on this insight, we present AEMatter, a matting network that is straightforward yet very effective. AEMatter adopts a Hybrid-Transformer backbone with appearance-enhanced axis-wise learning (AEAL) blocks to build a basic network with strong context aggregation learning capability. Furthermore, AEMatter leverages a large image training strategy to assist the network in learning context aggregation from data. Extensive experiments on five popular matting datasets demonstrate that the proposed AEMatter outperforms state-of-the-art matting methods by a large margin.

cs.CV↗

arXiv2024

Pseudogap phase as fluctuating pair density wave

The physical nature of pseudogap phase is one of the most important and intriguing problems towards understanding the key mechanism of high temperature superconductivity in cuprates. Theoretically, the square-lattice $t$-$J$ model is widely believed to be the simplest toy model that captures the essential physics of cuprate superconductors. We employ the Grassmann tensor product state approach to investigate uniform states in the underdoped ($δ\lesssim 0.1$) region. In addition to the previously known uniform $d$-wave state, we discover a strongly fluctuating pair density wave (PDW) state with wave vector $Q = (π, π)$. This fluctuating PDW state weakly breaks the $C_4$ rotational symmetry of the square lattice and has a lower or comparable energy to the $d$-wave state (depending on doping and the $t/J$ ratio), making it a promising candidate state for describing the pseudogap phase.

cond-mat.str-el↗

arXiv2024

MEGA-DAgger: Imitation Learning with Multiple Imperfect Experts

Imitation learning has been widely applied to various autonomous systems thanks to recent development in interactive algorithms that address covariate shift and compounding errors induced by traditional approaches like behavior cloning. However, existing interactive imitation learning methods assume access to one perfect expert. Whereas in reality, it is more likely to have multiple imperfect experts instead. In this paper, we propose MEGA-DAgger, a new DAgger variant that is suitable for interactive learning with multiple imperfect experts. First, unsafe demonstrations are filtered while aggregating the training data, so the imperfect demonstrations have little influence when training the novice policy. Next, experts are evaluated and compared on scenarios-specific metrics to resolve the conflicted labels among experts. Through experiments in autonomous racing scenarios, we demonstrate that policy learned using MEGA-DAgger can outperform both experts and policies learned using the state-of-the-art interactive imitation learning algorithms such as Human-Gated DAgger. The supplementary video can be found at \url{https://youtu.be/wPCht31MHrw}.

cs.LG↗

arXiv2024

Chebyshev pseudosite matrix product state approach for cluster perturbation theory

We introduce the Chebyshev pseudosite matrix product state approach (ChePSMPS) as a solver for cluster perturbation theory (CPT), crucial for simulating spectral functions in two-dimensional electron-phonon ($e$-ph) coupling systems. ChePSMPS distinguishes itself from conventional exact diagonalization solvers by supporting larger clusters, thereby significantly mitigating finite-size effects. Free from the fermion sign problem, ChePSMPS enhances its ability to explore $e$-ph effects and generate high-resolution spectral functions in doped Mott insulators. We use this method to simulate the spectra for both one- and two-dimensional Hubbard-Holstein models, highlighting its superiority over other methods. Our findings validate ChePSMPS as a powerful and reliable Green's function solver. In conjunction with embedding methods, ChePSMPS emerges as an essential tool for simulating strongly correlated $e$-ph coupling systems.

cond-mat.str-el↗

arXiv2024

ELODI: Ensemble Logit Difference Inhibition for Positive-Congruent Training

Negative flips are errors introduced in a classification system when a legacy model is updated. Existing methods to reduce the negative flip rate (NFR) either do so at the expense of overall accuracy by forcing a new model to imitate the old models, or use ensembles, which multiply inference cost prohibitively. We analyze the role of ensembles in reducing NFR and observe that they remove negative flips that are typically not close to the decision boundary, but often exhibit large deviations in the distance among their logits. Based on the observation, we present a method, called Ensemble Logit Difference Inhibition (ELODI), to train a classification system that achieves paragon performance in both error rate and NFR, at the inference cost of a single model. The method distills a homogeneous ensemble to a single student model which is used to update the classification system. ELODI also introduces a generalized distillation objective, Logit Difference Inhibition (LDI), which only penalizes the logit difference of a subset of classes with the highest logit values. On multiple image classification benchmarks, model updates with ELODI demonstrate superior accuracy retention and NFR reduction.

cs.LG↗

arXiv2024

The Physics potential of the CEPC. Prepared for the US Snowmass Community Planning Exercise (Snowmass 2021)

The Circular Electron Positron Collider (CEPC) is a large-scale collider facility that can serve as a factory of the Higgs, Z, and W bosons and is upgradable to run at the ttbar threshold. This document describes the latest CEPC nominal operation scenario and particle yields and updates the corresponding physics potential. A new detector concept is also briefly described. This submission is for consideration by the Snowmass process.

hep-ph↗

arXiv2024

Global phase diagram of doped quantum spin liquid on the Kagome lattice

It has long been believed that doped quantum spin liquids (QSLs) can give rise to fascinating quantum phases, including the possibility of high-temperature superconductivity (SC) as proposed by P. W. Anderson's resonating valence bond (RVB) scenario. The Kagome lattice $t$-$J$ model is known to exhibit spin liquid behavior at half-filling, making it an ideal system for studying the properties of doped QSL. In this study, we employ the fermionic projected entangled simplex state (PESS) method to investigate the ground state properties of the Kagome lattice $t$-$J$ model with $t/J = 3.0$. Our results reveal a phase transition from charge density wave (CDW) states to uniform states around a critical doping level $δ_c \approx 0.27$. Within the CDW phase, we observe different types of Wigner crystal (WC) formulated by doped holes that are energetically favored. As we enter the uniform phase, a non-Fermi liquid (NFL) state emerges within the doping range $0.27 < δ< 0.32$, characterized by an exponential decay of all correlation functions. With further hole doping, we discover the appearance of a pair density wave (PDW) state within a narrow doping region $0.32 < δ< 1/3$. We also discuss the potential experimental implications of our findings.

cond-mat.str-el↗

arXiv2024

Uncover the Premeditated Attacks: Detecting Exploitable Reentrancy Vulnerabilities by Identifying Attacker Contracts

Reentrancy, a notorious vulnerability in smart contracts, has led to millions of dollars in financial loss. However, current smart contract vulnerability detection tools suffer from a high false positive rate in identifying contracts with reentrancy vulnerabilities. Moreover, only a small portion of the detected reentrant contracts can actually be exploited by hackers, making these tools less effective in securing the Ethereum ecosystem in practice. In this paper, we propose BlockWatchdog, a tool that focuses on detecting reentrancy vulnerabilities by identifying attacker contracts. These attacker contracts are deployed by hackers to exploit vulnerable contracts automatically. By focusing on attacker contracts, BlockWatchdog effectively detects truly exploitable reentrancy vulnerabilities by identifying reentrant call flow. Additionally, BlockWatchdog is capable of detecting new types of reentrancy vulnerabilities caused by poor designs when using ERC tokens or user-defined interfaces, which cannot be detected by current rule-based tools. We implement BlockWatchdog using cross-contract static dataflow techniques based on attack logic obtained from an empirical study that analyzes attacker contracts from 281 attack incidents. BlockWatchdog is evaluated on 421,889 Ethereum contract bytecodes and identifies 113 attacker contracts that target 159 victim contracts, leading to the theft of Ether and tokens valued at approximately 908.6 million USD. Notably, only 18 of the identified 159 victim contracts can be reported by current reentrancy detection tools.

cs.CR↗

arXiv2024

Reduced Membrane Model for Liquid Crystal Polymer Networks: Asymptotics and Computation

We examine a reduced membrane model of liquid crystal polymer networks (LCNs) via asymptotics and computation. This model requires solving a minimization problem for a non-convex stretching energy. We show a formal asymptotic derivation of the 2D membrane model from 3D rubber elasticity. We construct approximate solutions with point defects. We design a finite element method with regularization, and propose a nonlinear gradient flow with Newton inner iteration to solve the non-convex discrete minimization problem. We present numerical simulations of practical interests to illustrate the ability of the model and our method to capture rich physical phenomena.

math.NA↗

arXiv2024

Is Crowdsourcing Breaking Your Bank? Cost-Effective Fine-Tuning of Pre-trained Language Models with Proximal Policy Optimization

Wide usage of ChatGPT has highlighted the potential of reinforcement learning from human feedback. However, its training pipeline relies on manual ranking, a resource-intensive process. To reduce labor costs, we propose a self-supervised text ranking approach for applying Proximal-Policy-Optimization to fine-tune language models while eliminating the need for human annotators. Our method begins with probabilistic sampling to encourage a language model to generate diverse responses for each input. We then employ TextRank and ISODATA algorithms to rank and cluster these responses based on their semantics. Subsequently, we construct a reward model to learn the rank and optimize our generative policy. Our experimental results, conducted using two language models on three tasks, demonstrate that the models trained by our method considerably outperform baselines regarding BLEU, GLEU, and METEOR scores. Furthermore, our manual evaluation shows that our ranking results exhibit a remarkably high consistency with that of humans. This research significantly reduces training costs of proximal policy-guided models and demonstrates the potential for self-correction of language models.

cs.CL↗

arXiv2024

Data-free Multi-label Image Recognition via LLM-powered Prompt Tuning

This paper proposes a novel framework for multi-label image recognition without any training data, called data-free framework, which uses knowledge of pre-trained Large Language Model (LLM) to learn prompts to adapt pretrained Vision-Language Model (VLM) like CLIP to multilabel classification. Through asking LLM by well-designed questions, we acquire comprehensive knowledge about characteristics and contexts of objects, which provides valuable text descriptions for learning prompts. Then we propose a hierarchical prompt learning method by taking the multi-label dependency into consideration, wherein a subset of category-specific prompt tokens are shared when the corresponding objects exhibit similar attributes or are more likely to co-occur. Benefiting from the remarkable alignment between visual and linguistic semantics of CLIP, the hierarchical prompts learned from text descriptions are applied to perform classification of images during inference. Our framework presents a new way to explore the synergies between multiple pre-trained models for novel category recognition. Extensive experiments on three public datasets (MS-COCO, VOC2007, and NUS-WIDE) demonstrate that our method achieves better results than the state-of-the-art methods, especially outperforming the zero-shot multi-label recognition methods by 4.7% in mAP on MS-COCO.

cs.CV↗

arXiv2024

MENTOR: Multi-level Self-supervised Learning for Multimodal Recommendation

With the increasing multimedia information, multimodal recommendation has received extensive attention. It utilizes multimodal information to alleviate the data sparsity problem in recommendation systems, thus improving recommendation accuracy. However, the reliance on labeled data severely limits the performance of multimodal recommendation models. Recently, self-supervised learning has been used in multimodal recommendations to mitigate the label sparsity problem. Nevertheless, the state-of-the-art methods cannot avoid the modality noise when aligning multimodal information due to the large differences in the distributions of different modalities. To this end, we propose a Multi-level sElf-supervised learNing for mulTimOdal Recommendation (MENTOR) method to address the label sparsity problem and the modality alignment problem. Specifically, MENTOR first enhances the specific features of each modality using the graph convolutional network (GCN) and fuses the visual and textual modalities. It then enhances the item representation via the item semantic graph for all modalities, including the fused modality. Then, it introduces two multilevel self-supervised tasks: the multilevel cross-modal alignment task and the general feature enhancement task. The multilevel cross-modal alignment task aligns each modality under the guidance of the ID embedding from multiple levels while maintaining the historical interaction information. The general feature enhancement task enhances the general feature from both the graph and feature perspectives to improve the robustness of our model. Extensive experiments on three publicly available datasets demonstrate the effectiveness of our method. Our code is publicly available at https://github.com/Jinfeng-Xu/MENTOR.

cs.IR↗

arXiv2023

ROS package search for robot software development: a knowledge graph-based approach

ROS (Robot Operating System) packages have become increasingly popular as a type of software artifact that can be effectively reused in robotic software development. Indeed, finding suitable ROS packages that closely match the software's functional requirements from the vast number of available packages is a nontrivial task using current search methods. The traditional search methods for ROS packages often involve inputting keywords related to robotic tasks into general-purpose search engines or code hosting platforms to obtain approximate results of all potentially suitable ROS packages. However, the accuracy of these search methods remains relatively low because the task-related keywords may not precisely match the functionalities offered by the ROS packages. To improve the search accuracy of ROS packages, this paper presents a novel semantic-based search approach that relies on the semantic-level ROS Package Knowledge Graph (RPKG) to automatically retrieve the most suitable ROS packages. Firstly, to construct the RPKG, we employ multi-dimensional feature extraction techniques to extract semantic concepts from the dataset of ROS package text descriptions. The semantic features extracted from this process result in a substantial number of entities and relationships. Subsequently, we create a robot domain-specific small corpus and further fine-tune a pre-trained language model, BERT-ROS, to generate embeddings that effectively represent the semantics of the extracted features. These embeddings play a crucial role in facilitating semantic-level understanding and comparisons during the ROS package search process within the RPKG. Secondly, we introduce a novel semantic matching-based search algorithm that incorporates the weighted similarities of multiple features from user search queries, which searches out more accurate ROS packages than the traditional keyword search method.

cs.SE↗

arXiv2023

Not All Negatives Are Worth Attending to: Meta-Bootstrapping Negative Sampling Framework for Link Prediction

The rapid development of graph neural networks (GNNs) encourages the rising of link prediction, achieving promising performance with various applications. Unfortunately, through a comprehensive analysis, we surprisingly find that current link predictors with dynamic negative samplers (DNSs) suffer from the migration phenomenon between "easy" and "hard" samples, which goes against the preference of DNS of choosing "hard" negatives, thus severely hindering capability. Towards this end, we propose the MeBNS framework, serving as a general plugin that can potentially improve current negative sampling based link predictors. In particular, we elaborately devise a Meta-learning Supported Teacher-student GNN (MST-GNN) that is not only built upon teacher-student architecture for alleviating the migration between "easy" and "hard" samples but also equipped with a meta learning based sample re-weighting module for helping the student GNN distinguish "hard" samples in a fine-grained manner. To effectively guide the learning of MST-GNN, we prepare a Structure enhanced Training Data Generator (STD-Generator) and an Uncertainty based Meta Data Collector (UMD-Collector) for supporting the teacher and student GNN, respectively. Extensive experiments show that the MeBNS achieves remarkable performance across six link prediction benchmark datasets.

cs.LG↗