Instruction tuning is an effective technique to align large language models (LLMs) with human intents. In this work, we investigate how an adversary can exploit instruction tuning by injecting specific instruction-following examples into the training data that intentionally changes the model's behavior. For example, an adversary can achieve content injection by injecting training examples that mention target content and eliciting such behavior from downstream models. To achieve this goal, we propose \textit{AutoPoison}, an automated data poisoning pipeline. It naturally and coherently incorporates versatile attack goals into poisoned data with the help of an oracle LLM. We showcase two example attacks: content injection and over-refusal attacks, each aiming to induce a specific exploitable behavior. We quantify and benchmark the strength and the stealthiness of our data poisoning scheme. Our results show that AutoPoison allows an adversary to change a model's behavior by poisoning only a small fraction of data while maintaining a high level of stealthiness in the poisoned examples. We hope our work sheds light on how data quality affects the behavior of instruction-tuned models and raises awareness of the importance of data quality for responsible deployments of LLMs. Code is available at \url{//github.com/azshue/AutoPoison}.
相關內容
We report some results regarding the mechanization of normative (preference-based) conditional reasoning. Our focus is on Aqvist's system E for conditional obligation (and its extensions). Our mechanization is achieved via a shallow semantical embedding in Isabelle/HOL. We consider two possible uses of the framework. The first one is as a tool for meta-reasoning about the considered logic. We employ it for the automated verification of deontic correspondences (broadly conceived) and related matters, analogous to what has been previously achieved for the modal logic cube. The second use is as a tool for assessing ethical arguments. We provide a computer encoding of a well-known paradox in population ethics, Parfit's repugnant conclusion. Whether the presented encoding increases or decreases the attractiveness and persuasiveness of the repugnant conclusion is a question we would like to pass on to philosophy and ethics.
In this work, we present an abstract theory for the approximation of operator-valued Riccati equations posed on Hilbert spaces. It is demonstrated here, under the assumption of compactness in the coefficient operators, that the error of the approximate solution to the operator-valued Riccati equation is bounded above by the approximation error of the governing semigroup. One significant outcome of this result is the correct prediction of optimal convergence for finite element approximations of the operator-valued Riccati equations for when the governing semigroup involves parabolic, as well as hyperbolic processes. We derive the abstract theory for the time-dependent and time-independent operator-valued Riccati equations in the first part of this work. In the second part, we prove optimal convergence rates for the finite element approximation of the functional gain associated with model one-dimensional weakly damped wave and thermal LQR control systems. These theoretical claims are then corroborated with computational evidence.
Backdoor (Trojan) attacks are an important type of adversarial exploit against deep neural networks (DNNs), wherein a test instance is (mis)classified to the attacker's target class whenever the attacker's backdoor trigger is present. In this paper, we reveal and analyze an important property of backdoor attacks: a successful attack causes an alteration in the distribution of internal layer activations for backdoor-trigger instances, compared to that for clean instances. Even more importantly, we find that instances with the backdoor trigger will be correctly classified to their original source classes if this distribution alteration is corrected. Based on our observations, we propose an efficient and effective method that achieves post-training backdoor mitigation by correcting the distribution alteration using reverse-engineered triggers. Notably, our method does not change any trainable parameters of the DNN, but achieves generally better mitigation performance than existing methods that do require intensive DNN parameter tuning. It also efficiently detects test instances with the trigger, which may help to catch adversarial entities in the act of exploiting the backdoor.
Phase information has a significant impact on speech perceptual quality and intelligibility. However, existing speech enhancement methods encounter limitations in explicit phase estimation due to the non-structural nature and wrapping characteristics of the phase, leading to a bottleneck in enhanced speech quality. To overcome the above issue, in this paper, we proposed MP-SENet, a novel Speech Enhancement Network which explicitly enhances Magnitude and Phase spectra in parallel. The proposed MP-SENet adopts a codec architecture in which the encoder and decoder are bridged by time-frequency Transformers along both time and frequency dimensions. The encoder aims to encode time-frequency representations derived from the input distorted magnitude and phase spectra. The decoder comprises dual-stream magnitude and phase decoders, directly enhancing magnitude and wrapped phase spectra by incorporating a magnitude estimation architecture and a phase parallel estimation architecture, respectively. To train the MP-SENet model effectively, we define multi-level loss functions, including mean square error and perceptual metric loss of magnitude spectra, anti-wrapping loss of phase spectra, as well as mean square error and consistency loss of short-time complex spectra. Experimental results demonstrate that our proposed MP-SENet excels in high-quality speech enhancement across multiple tasks, including speech denoising, dereverberation, and bandwidth extension. Compared to existing phase-aware speech enhancement methods, it successfully avoids the bidirectional compensation effect between the magnitude and phase, leading to a better harmonic restoration. Notably, for the speech denoising task, the MP-SENet yields a state-of-the-art performance with a PESQ of 3.60 on the public VoiceBank+DEMAND dataset.
When is heterogeneity in the composition of an autonomous robotic team beneficial and when is it detrimental? We investigate and answer this question in the context of a minimally viable model that examines the role of heterogeneous speeds in perimeter defense problems, where defenders share a total allocated speed budget. We consider two distinct problem settings and develop strategies based on dynamic programming and on local interaction rules. We present a theoretical analysis of both approaches and our results are extensively validated using simulations. Interestingly, our results demonstrate that the viability of heterogeneous teams depends on the amount of information available to the defenders. Moreover, our results suggest a universality property: across a wide range of problem parameters the optimal ratio of the speeds of the defenders remains nearly constant.
In the past decade, we have witnessed the rise of deep learning to dominate the field of artificial intelligence. Advances in artificial neural networks alongside corresponding advances in hardware accelerators with large memory capacity, together with the availability of large datasets enabled researchers and practitioners alike to train and deploy sophisticated neural network models that achieve state-of-the-art performance on tasks across several fields spanning computer vision, natural language processing, and reinforcement learning. However, as these neural networks become bigger, more complex, and more widely used, fundamental problems with current deep learning models become more apparent. State-of-the-art deep learning models are known to suffer from issues that range from poor robustness, inability to adapt to novel task settings, to requiring rigid and inflexible configuration assumptions. Ideas from collective intelligence, in particular concepts from complex systems such as self-organization, emergent behavior, swarm optimization, and cellular systems tend to produce solutions that are robust, adaptable, and have less rigid assumptions about the environment configuration. It is therefore natural to see these ideas incorporated into newer deep learning methods. In this review, we will provide a historical context of neural network research's involvement with complex systems, and highlight several active areas in modern deep learning research that incorporate the principles of collective intelligence to advance its current capabilities. To facilitate a bi-directional flow of ideas, we also discuss work that utilize modern deep learning models to help advance complex systems research. We hope this review can serve as a bridge between complex systems and deep learning communities to facilitate the cross pollination of ideas and foster new collaborations across disciplines.
AI is undergoing a paradigm shift with the rise of models (e.g., BERT, DALL-E, GPT-3) that are trained on broad data at scale and are adaptable to a wide range of downstream tasks. We call these models foundation models to underscore their critically central yet incomplete character. This report provides a thorough account of the opportunities and risks of foundation models, ranging from their capabilities (e.g., language, vision, robotics, reasoning, human interaction) and technical principles(e.g., model architectures, training procedures, data, systems, security, evaluation, theory) to their applications (e.g., law, healthcare, education) and societal impact (e.g., inequity, misuse, economic and environmental impact, legal and ethical considerations). Though foundation models are based on standard deep learning and transfer learning, their scale results in new emergent capabilities,and their effectiveness across so many tasks incentivizes homogenization. Homogenization provides powerful leverage but demands caution, as the defects of the foundation model are inherited by all the adapted models downstream. Despite the impending widespread deployment of foundation models, we currently lack a clear understanding of how they work, when they fail, and what they are even capable of due to their emergent properties. To tackle these questions, we believe much of the critical research on foundation models will require deep interdisciplinary collaboration commensurate with their fundamentally sociotechnical nature.
Recent years have seen important advances in the quality of state-of-the-art models, but this has come at the expense of models becoming less interpretable. This survey presents an overview of the current state of Explainable AI (XAI), considered within the domain of Natural Language Processing (NLP). We discuss the main categorization of explanations, as well as the various ways explanations can be arrived at and visualized. We detail the operations and explainability techniques currently available for generating explanations for NLP model predictions, to serve as a resource for model developers in the community. Finally, we point out the current gaps and encourage directions for future work in this important research area.
Deep learning has penetrated all aspects of our lives and brought us great convenience. However, the process of building a high-quality deep learning system for a specific task is not only time-consuming but also requires lots of resources and relies on human expertise, which hinders the development of deep learning in both industry and academia. To alleviate this problem, a growing number of research projects focus on automated machine learning (AutoML). In this paper, we provide a comprehensive and up-to-date study on the state-of-the-art AutoML. First, we introduce the AutoML techniques in details according to the machine learning pipeline. Then we summarize existing Neural Architecture Search (NAS) research, which is one of the most popular topics in AutoML. We also compare the models generated by NAS algorithms with those human-designed models. Finally, we present several open problems for future research.
Machine learning techniques have deeply rooted in our everyday life. However, since it is knowledge- and labor-intensive to pursue good learning performance, human experts are heavily involved in every aspect of machine learning. In order to make machine learning techniques easier to apply and reduce the demand for experienced human experts, automated machine learning (AutoML) has emerged as a hot topic with both industrial and academic interest. In this paper, we provide an up to date survey on AutoML. First, we introduce and define the AutoML problem, with inspiration from both realms of automation and machine learning. Then, we propose a general AutoML framework that not only covers most existing approaches to date but also can guide the design for new methods. Subsequently, we categorize and review the existing works from two aspects, i.e., the problem setup and the employed techniques. Finally, we provide a detailed analysis of AutoML approaches and explain the reasons underneath their successful applications. We hope this survey can serve as not only an insightful guideline for AutoML beginners but also an inspiration for future research.
注冊地址： 北京市海淀區羊坊店路18號2幢3層301-191