检索结果-内蒙古大学图书馆

12th International Conference on Learning Representations, ICLR 2024

作者： Wang, Lean Yang, Wenkai Chen, Deli Zhou, Hao Lin, Yankai Meng, Fandong Zhou, Jie Sun, Xu National Key Laboratory for Multimedia Information Processing School of Computer Science Peking University China Gaoling School of Artificial Intelligence Renmin University of China China Pattern Recognition Center WeChat AI Tencent Inc. China DeepSeek AI China

As large language models (LLMs) generate texts with increasing fluency and realism, there is a growing need to identify the source of texts to prevent the abuse of LLMs. Text watermarking techniques have proven reliable in distinguishing whether a text is generated by LLMs by injecting hidden patterns. However, we argue that existing LLM watermarking methods are encoding-inefficient and cannot flexibly meet the diverse information encoding needs (such as encoding model version, generation time, user id, etc.). In this work, we conduct the first systematic study on the topic of Codable Text Watermarking for LLMs (CTWL) that allows text watermarks to carry multi-bit customizable information. First of all, we study the taxonomy of LLM watermarking technologies and give a mathematical formulation for CTWL. Additionally, we provide a comprehensive evaluation system for CTWL: (1) watermarking success rate, (2) robustness against various corruptions, (3) coding rate of payload information, (4) encoding and decoding efficiency, (5) impacts on the quality of the generated text. To meet the requirements of these non-Pareto-improving metrics, we follow the most prominent vocabulary partition-based watermarking direction, and devise an advanced CTWL method named Balance-Marking. The core idea of our method is to use a proxy language model to split the vocabulary into probability-balanced parts, thereby effectively maintaining the quality of the watermarked text. Extensive experimental results show that our method outperforms the baseline under comprehensive evaluation. Our code is available at https://***/lancopku/codable-watermarking-for-llm. © 2024 12th International Conference on Learning Representations, ICLR 2024. All rights reserved.

关键词： Encoding (symbols)

来源：评论

学校读者我要写书评

暂无评论

Watch Out for Your Agents! Investigating Backdoor Threats to LLM-Based Agents 38

Watch Out for Your Agents! Investigating Backdoor Threats to...

引用

38th Conference on Neural Information processing Systems, NeurIPS 2024

作者： Yang, Wenkai Bi, Xiaohan Lin, Yankai Chen, Sishuo Zhou, Jie Sun, Xu Gaoling School of Artificial Intelligence Renmin University of China Beijing China Center for Data Science Peking University China Pattern Recognition Center WeChat AI Tencent Inc. China National Key Laboratory for Multimedia Information Processing School of Computer Science Peking University China

Driven by the rapid development of Large Language Models (LLMs), LLM-based agents have been developed to handle various real-world applications, including finance, healthcare, and shopping, etc. It is crucial to ensure the reliability and security of LLM-based agents during applications. However, the safety issues of LLM-based agents are currently under-explored. In this work, we take the first step to investigate one of the typical safety threats, backdoor attack, to LLM-based agents. We first formulate a general framework of agent backdoor attacks, then we present a thorough analysis of different forms of agent backdoor attacks. Specifically, compared with traditional backdoor attacks on LLMs that are only able to manipulate the user inputs and model outputs, agent backdoor attacks exhibit more diverse and covert forms: (1) From the perspective of the final attacking outcomes, the agent backdoor attacker can not only choose to manipulate the final output distribution, but also introduce the malicious behavior in an intermediate reasoning step only, while keeping the final output correct. (2) Furthermore, the former category can be divided into two subcategories based on trigger locations, in which the backdoor trigger can either be hidden in the user query or appear in an intermediate observation returned by the external environment. We implement the above variations of agent backdoor attacks on two typical agent tasks including web shopping and tool utilization. Extensive experiments show that LLM-based agents suffer severely from backdoor attacks and such backdoor vulnerability cannot be easily mitigated by current textual backdoor defense algorithms. This indicates an urgent need for further research on the development of targeted defenses against backdoor attacks on LLM-based agents. © 2024 Neural information processing systems foundation. All rights reserved.

关键词：

来源：评论

学校读者我要写书评

暂无评论

A Unified Framework for Multi-Intent Spoken Language Understanding with Prompting

A Unified Framework for Multi-Intent Spoken Language Underst...

引用

International Conference on Acoustics, Speech, and signal processing (ICASSP)

作者： Feifan Song Lianzhe Huang Houfeng Wang School of Computer Science Peking University National Key Laboratory of Multimedia Information Processing Peking University Pattern Recognition Center WeChat AI Tencent

ChatGPT has demonstrated impressive capabilities in building conversations. However, for Spoken Language Understanding (SLU) with multiple intents, traditional approaches where Intent Detection and Slot Filling are jointly modeled with distinct formulations hinder networks from effectively extracting shared features. In this work, we describe a Prompt-based SLU (PromptSLU) framework, to intuitively unify two sub-tasks into the same form for a common pre-trained model. Specifically, variable intents are predicted first, then naturally embedded into prompts to guide slot-value inference from a semantic perspective. Furthermore, we are inspired by multi-task learning to introduce an auxiliary sub-task and a concise general objective, which helps to learn relationships among provided labels. Experiment results show that our framework outperforms several competitive baselines on two datasets. The source code is available at https://***/F2-Song/PromptSLU.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Fed-FA: theoretically modeling client data divergence for federated language backdoor defense 23

Fed-FA: theoretically modeling client data divergence for fe...

引用

Proceedings of the 37th International Conference on Neural Information processing Systems

作者： Zhiyuan Zhang Deli Chen Hao Zhou Fandong Meng Jie Zhou Xu Sun National Key Laboratory for Multimedia Information Processing School of Computer Science Peking University and Pattern Recognition Center WeChat AI Tencent Inc. China Pattern Recognition Center WeChat AI Tencent Inc. China National Key Laboratory for Multimedia Information Processing School of Computer Science Peking University

Federated learning algorithms enable neural network models to be trained across multiple decentralized edge devices without sharing private data. However, they are susceptible to backdoor attacks launched by malicious clients. Existing robust federated aggregation algorithms heuristically detect and exclude suspicious clients based on their parameter distances, but they are ineffective on Natural Language processing (NLP) tasks. The main reason is that, although text backdoor patterns are obvious at the underlying dataset level, they are usually hidden at the parameter level, since injecting backdoors into texts with discrete feature space has less impact on the statistics of the model parameters. To settle this issue, we propose to identify backdoor clients by explicitly modeling the data divergence among clients in federated NLP systems. Through theoretical analysis, we derive the f-divergence indicator to estimate the client data divergence with aggregation updates and Hessians. Furthermore, we devise a dataset synthesization method with a Hessian reassignment mechanism guided by the diffusion theory to address the key challenge of inaccessible datasets in calculating clients' data Hessians. We then present the novel Federated F-Divergence-Based Aggregation (Fed-FA) algorithm, which leverages the f-divergence indicator to detect and discard suspicious clients. Extensive empirical results show that Fed-FA outperforms all the parameter distance-based methods in defending against backdoor attacks among various natural language backdoor attack scenarios.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Diffusion Theory as a Scalpel: Detecting and Purifying Poisonous Dimensions in Pre-trained Language Models Caused by Backdoor or Bias

arXiv

引用

arXiv 2023年

作者： Zhang, Zhiyuan Chen, Deli Zhou, Hao Meng, Fandong Zhou, Jie Sun, Xu National Key Laboratory for Multimedia Information Processing School of Computer Science Peking University China Pattern Recognition Center WeChat AI Tencent Inc. China

Pre-trained Language Models (PLMs) may be poisonous with backdoors or bias injected by the suspicious attacker during the fine-tuning process. A core challenge of purifying potentially poisonous PLMs is precisely finding poisonous dimensions. To settle this issue, we propose the Fine-purifying approach, which utilizes the diffusion theory to study the dynamic process of fine-tuning for finding potentially poisonous dimensions. According to the relationship between parameter drifts and Hessians of different dimensions, we can detect poisonous dimensions with abnormal dynamics, purify them by resetting them to clean pre-trained weights, and then fine-tune the purified weights on a small clean dataset. To the best of our knowledge, we are the first to study the dynamics guided by the diffusion theory for safety or defense purposes. Experimental results validate the effectiveness of Fine-purifying even with a small clean dataset. Copyright © 2023, The Authors. All rights reserved.

关键词： Diffusion

来源：评论

学校读者我要写书评

暂无评论

Label Words are Anchors: An Information Flow Perspective for Understanding In-Context Learning

arXiv

引用

arXiv 2023年

作者： Wang, Lean Li, Lei Dai, Damai Chen, Deli Zhou, Hao Meng, Fandong Zhou, Jie Sun, Xu National Key Laboratory for Multimedia Information Processing School of Computer Science Peking University China Pattern Recognition Center WeChat AI Tencent Inc. China

In-context learning (ICL) emerges as a promising capability of large language models (LLMs) by providing them with demonstration examples to perform diverse tasks. However, the underlying mechanism of how LLMs learn from the provided context remains under-explored. In this paper, we investigate the working mechanism of ICL through an information flow lens. Our findings reveal that label words in the demonstration examples function as anchors: (1) semantic information aggregates into label word representations during the shallow computation layers' processing;(2) the consolidated information in label words serves as a reference for LLMs' final predictions. Based on these insights, we introduce an anchor re-weighting method to improve ICL performance, a demonstration compression technique to expedite inference, and an analysis framework for diagnosing ICL errors in GPT2-XL. The promising applications of our findings again validate the uncovered ICL working mechanism and pave the way for future studies. © 2023, CC BY.

关键词： Demonstrations

来源：评论

学校读者我要写书评

暂无评论

Decentralized Kernel Ridge Regression Based on Data-Dependent Random Feature

arXiv

引用

arXiv 2024年

作者： Yang, Ruikai He, Fan He, Mingzhen Yang, Jie Huang, Xiaolin The Institute of Image Processing and Pattern Recognition The MOE Key Laboratory of System Control and Information Processing Shanghai Jiao Tong University Shanghai200240 China The STADIUS Center for Dynamical Systems Signal Processing and Data Analytics KU Leuven LeuvenB-3001 Belgium

Random feature (RF) has been widely used for node consistency in decentralized kernel ridge regression (KRR). Currently, the consistency is guaranteed by imposing constraints on coefficients of features, necessitating that the random features on different nodes are identical. However, in many applications, data on different nodes varies significantly on the number or distribution, which calls for adaptive and data-dependent methods that generate different RFs. To tackle the essential difficulty, we propose a new decentralized KRR algorithm that pursues consensus on decision functions, which allows great flexibility and well adapts data on nodes. The convergence is rigorously given and the effectiveness is numerically verified: by capturing the characteristics of the data on each node, while maintaining the same communication costs as other methods, we achieved an average regression accuracy improvement of 25.5% across six real-world data sets. Copyright © 2024, The Authors. All rights reserved.

关键词： Consensus algorithm

来源：评论

学校读者我要写书评

暂无评论

STRNet:Triple-stream Spatiotemporal Relation Network for Action recognition

引用

International Journal of Automation and computing 2021年第5期18卷 718-730页

作者： Zhi-Wei Xu Xiao-Jun Wu Josef Kittler School of Artificial Intelligence and Computer Science Jiangnan UniversityWuxi 214122China Jiangsu Provincial Engineering Laboratory of Pattern Recognition and Computational Intelligence Wuxi 214122China Centre for Vision Speech and Signal ProcessingUniversity of SurreyGuildfordGU27XHUK

Learning comprehensive spatiotemporal features is crucial for human action recognition. Existing methods tend to model the spatiotemporal feature blocks in an integrate-separate-integrate form, such as appearance-and-relation network(ARTNet) and spatiotemporal and motion network(STM). However, with blocks stacking up, the rear part of the network has poor interpretability. To avoid this problem, we propose a novel architecture called spatial temporal relation network(STRNet), which can learn explicit information of appearance, motion and especially the temporal relation information. Specifically, our STRNet is constructed by three branches,which separates the features into 1) appearance pathway, to obtain spatial semantics, 2) motion pathway, to reinforce the spatiotemporal feature representation, and 3) relation pathway, to focus on capturing temporal relation details of successive frames and to explore long-term representation dependency. In addition, our STRNet does not just simply merge the multi-branch information, but we apply a flexible and effective strategy to fuse the complementary information from multiple pathways. We evaluate our network on four major action recognition benchmarks: Kinetics-400, UCF-101, HMDB-51, and Something-Something v1, demonstrating that the performance of our STRNet achieves the state-of-the-art result on the UCF-101 and HMDB-51 datasets, as well as a comparable accuracy with the state-of-the-art method on Something-Something v1 and Kinetics-400.

关键词： Action recognition spatiotemporal relation multi-branch fusion long-term representation video classification

来源：评论

学校读者我要写书评

暂无评论

Learning Analysis of Kernel Ridgeless Regression with Asymmetric Kernel Learning

arXiv

引用

arXiv 2024年

作者： He, Fan He, Mingzhen Shi, Lei Huang, Xiaolin Suykens, Johan A.K. STADIUS Center for Dynamical Systems Signal Processing and Data Analytics KU Leuven Leuven Belgium MOE Key Laboratory of System Control and Information Processing Institute of Image Processing and Pattern Recognition Shanghai Jiao Tong University Shanghai China Shanghai Key Laboratory for Contemporary Applied Mathematics School of Mathematical Sciences Fudan University Shanghai200433 China Shanghai Artificial Intelligence Laboratory Shanghai200232 China MOE Key Laboratory of System Control and Information Processing Institute of Image Processing and Pattern Recognition Institute of Medical Robotics Shanghai Jiao Tong University Shanghai200240 China

Ridgeless regression has garnered attention among researchers, particularly in light of the "Benign Overfitting" phenomenon, where models interpolating noisy samples demonstrate robust generalization. However, kernel ridgeless regression does not always perform well due to the lack of flexibility. This paper enhances kernel ridgeless regression with Locally-Adaptive-Bandwidths (LAB) RBF kernels, incorporating kernel learning techniques to improve performance in both experiments and theory. For the first time, we demonstrate that functions learned from LAB RBF kernels belong to an integral space of Reproducible Kernel Hilbert Spaces (RKHSs). Despite the absence of explicit regularization in the proposed model, its optimization is equivalent to solving an 0-regularized problem in the integral space of RKHSs, elucidating the origin of its generalization ability. Taking an approximation analysis viewpoint, we introduce an lq-norm analysis technique (with 0 © 2024, CC BY.

关键词： Vector spaces

来源：评论

学校读者我要写书评

暂无评论

A Riemannian Residual Learning Mechanism for SPD Network

A Riemannian Residual Learning Mechanism for SPD Network

引用

International Joint Conference on Neural Networks (IJCNN)

作者： Zhenyu Cai Rui Wang Tianyang Xu Xiaojun Wu Josef Kittler School of Artificial Intelligence and Computer Science Jiangnan University Wuxi China Jiangsu Provincial Engineering Laboratory of Pattern Recognition and Computational Wuxi China Centre for Vision Speech and Signal Processing University of Surrey Guildford U.K.

ISBN: (数字)9798350359312

ISBN: (纸本)9798350359329

The generalization of Euclidean network paradigm to the Riemannian manifolds has attracted much attention for offering useful geometric representations in processing manifold-valued data in recent years. However, the information degradation during data compression mapping hinders Riemannian networks from going deeper, and there are very few solutions specifically designed for this problem. Given the remarkable success of deep Residual learning in Euclidean networks, a novel Riemannian residual learning mechanism (RRLM) is proposed in the context of Symmetric Positive Definite (SPD) manifolds, enabling the characterization of deep spatiotemporal features while preserving the manifold properties. Based on RRLM, a stack of SPD manifold-constrained residual-like blocks is designed on the tail of the original SPDNet(backbone) for the sake of conducting deep Riemannian residual learning. For simplicity, we refer to the network architecture introduced above as Riemannian residual SPD network (ResSPDNet). The experimental results achieved on three types of visual classification tasks, i.e., facial emotion recognition, drone recognition, and action recognition, demonstrate that our method can achieve improved accuracy with a deepened network structure.

关键词： Manifolds Learning systems Degradation Emotion recognition Visualization Accuracy Face recognition

来源：评论

学校读者我要写书评

暂无评论

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案：

请选择收藏分类：

通借通还

建议与咨询 留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案： 新增检索档案 确定 取消

请选择收藏分类： 新增自定义分类 确定 取消

通借通还

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

请选择保存的检索档案：

请选择收藏分类：