Industry practitioners care about small improvements in malware detection accuracy because their models are deployed to hundreds of millions of machines, meaning a 0.1\% change can cause an overwhelming number of false positives. However, academic research is often restrained to public datasets on the order of ten thousand samples and is too small to detect improvements that may be relevant to industry. Working within these constraints, we devise an approach to generate a benchmark of configurable difficulty from a pool of available samples. This is done by leveraging malware family information from tools like AVClass to construct training/test splits that have different generalization rates, as measured by a secondary model. Our experiments will demonstrate that using a less accurate secondary model with disparate features is effective at producing benchmarks for a more sophisticated target model that is under evaluation. We also ablate against alternative designs to show the need for our approach.
相關內容
ACM/IEEE第23屆模型驅動工程語言和系統國際會議，是模型驅動軟件和系統工程的首要會議系列，由ACM-SIGSOFT和IEEE-TCSE支持組織。自1998年以來，模型涵蓋了建模的各個方面，從語言和方法到工具和應用程序。模特的參加者來自不同的背景，包括研究人員、學者、工程師和工業專業人士。MODELS 2019是一個論壇，參與者可以圍繞建模和模型驅動的軟件和系統交流前沿研究成果和創新實踐經驗。今年的版本將為建模社區提供進一步推進建模基礎的機會，并在網絡物理系統、嵌入式系統、社會技術系統、云計算、大數據、機器學習、安全、開源等新興領域提出建模的創新應用以及可持續性。
官網鏈接： ·
前向
·
論文
·
軟件工程
·
For many decades, formal methods are considered to be the way forward to help the software industry to make more reliable and trustworthy software. However, despite this strong belief and many individual success stories, no real change in industrial software development seems to be occurring. In fact, the software industry itself is moving forward rapidly, and the gap between what formal methods can achieve and the daily software-development practice does not appear to be getting smaller (and might even be growing). In the past, many recommendations have already been made on how to develop formal-methods research in order to close this gap. This paper investigates why the gap nevertheless still exists and provides its own recommendations on what can be done by the formal-methods-research community to bridge it. Our recommendations do not focus on open research questions. In fact, formal-methods tools and techniques are already of high quality and can address many non-trivial problems; we do give some technical recommendations on how tools and techniques can be made more accessible. To a greater extent, we focus on the human aspect: how to achieve impact, how to change the way of thinking of the various stakeholders about this issue, and in particular, as a research community, how to alter our behaviour, and instead of competing, collaborate to address this issue.
Membership inference attacks (MIAs) attempt to predict whether a particular datapoint is a member of a target model's training data. Despite extensive research on traditional machine learning models, there has been limited work studying MIA on the pre-training data of large language models (LLMs). We perform a large-scale evaluation of MIAs over a suite of language models (LMs) trained on the Pile, ranging from 160M to 12B parameters. We find that MIAs barely outperform random guessing for most settings across varying LLM sizes and domains. Our further analyses reveal that this poor performance can be attributed to (1) the combination of a large dataset and few training iterations, and (2) an inherently fuzzy boundary between members and non-members. We identify specific settings where LLMs have been shown to be vulnerable to membership inference and show that the apparent success in such settings can be attributed to a distribution shift, such as when members and non-members are drawn from the seemingly identical domain but with different temporal ranges. We release our code and data as a unified benchmark package that includes all existing MIAs, supporting future work.
The value-loading problem is a significant challenge for researchers aiming to create artificial intelligence (AI) systems that align with human values and preferences. This problem requires a method to define and regulate safe and optimal limits of AI behaviors. In this work, we propose HALO (Hormetic ALignment via Opponent processes), a regulatory paradigm that uses hormetic analysis to regulate the behavioral patterns of AI. Behavioral hormesis is a phenomenon where low frequencies of a behavior have beneficial effects, while high frequencies are harmful. By modeling behaviors as allostatic opponent processes, we can use either Behavioral Frequency Response Analysis (BFRA) or Behavioral Count Response Analysis (BCRA) to quantify the hormetic limits of repeatable behaviors. We demonstrate how HALO can solve the 'paperclip maximizer' scenario, a thought experiment where an unregulated AI tasked with making paperclips could end up converting all matter in the universe into paperclips. Our approach may be used to help create an evolving database of 'values' based on the hedonic calculus of repeatable behaviors with decreasing marginal utility. This positions HALO as a promising solution for the value-loading problem, which involves embedding human-aligned values into an AI system, and the weak-to-strong generalization problem, which explores whether weak models can supervise stronger models as they become more intelligent. Hence, HALO opens several research avenues that may lead to the development of a computational value system that allows an AI algorithm to learn whether the decisions it makes are right or wrong.
Manipulative design in user interfaces (conceptualized as dark patterns) has emerged as a significant impediment to the ethical design of technology and a threat to user agency and freedom of choice. While previous research focused on exploring these patterns in the context of graphical user interfaces, the impact of speech has largely been overlooked. We conducted a listening test (N = 50) to elicit participants' preferences regarding different synthetic voices that varied in terms of synthesis method (concatenative vs. neural) and prosodic qualities (speech pace and pitch variance), and then evaluated their impact in an online decision-making study (N = 101). Our results indicate a significant effect of voice qualities on the participant's choices, independently from the content of the available options. Our results also indicate that the voice's perceived engagement, ease of understanding, and domain fit directly translate to its impact on participants' behaviour in decision-making tasks.
Writing declarative models has numerous benefits, ranging from automated reasoning and correction of design-level properties before systems are built, to automated testing and debugging of their implementations after they are built. Alloy is a declarative modeling language that is well-suited for verifying system designs. A key strength of Alloy is its scenario-finding toolset, the Analyzer, which allows users to explore all valid scenarios that adhere to the model's constraints up to a user-provided scope. However, even with visualized scenarios, it is difficult to write correct Alloy models. To address this, a growing body of work explores different techniques for debugging Alloy models. In order to develop and evaluate these techniques in an effective manor, this paper presents an empirical study of over 97,000 models written by novice users trying to learn Alloy. We investigate how users write both correct and incorrect models in order to produce a comprehensive benchmark for future use as well as a series of observations to guide debugging and educational efforts for Alloy model development.
Thoughtfully designing services and rigorously testing software to support personal information management (PIM) requires understanding the relevant collections, but relatively little is known about what people keep in their file collections, especially personal collections. Complementing recent work on the structure of 348 file collections, we examine those collections' contents, how much content is duplicated, and how collections used for personal matters differ from those used for study and work. Though all collections contain many images, some intuitively common file types are surprisingly scarce. Personal collections contain more audio than others, knowledge workers' collections contain more text documents but far fewer folders, and IT collections exhibit unusual traits. Collection duplication is correlated to collections' structural traits, but surprisingly, not to collection age. We discuss our findings in light of prior works and provide implications for various kinds of information research.
Diffusion models have emerged as a prominent class of generative models, surpassing previous methods regarding sample quality and training stability. Recent works have shown the advantages of diffusion models in improving reinforcement learning (RL) solutions, including as trajectory planners, expressive policy classes, data synthesizers, etc. This survey aims to provide an overview of the advancements in this emerging field and hopes to inspire new avenues of research. First, we examine several challenges encountered by current RL algorithms. Then, we present a taxonomy of existing methods based on the roles played by diffusion models in RL and explore how the existing challenges are addressed. We further outline successful applications of diffusion models in various RL-related tasks while discussing the limitations of current approaches. Finally, we conclude the survey and offer insights into future research directions, focusing on enhancing model performance and applying diffusion models to broader tasks. We are actively maintaining a GitHub repository for papers and other related resources in applying diffusion models in RL: //github.com/apexrl/Diff4RLSurvey .
Non-convex optimization is ubiquitous in modern machine learning. Researchers devise non-convex objective functions and optimize them using off-the-shelf optimizers such as stochastic gradient descent and its variants, which leverage the local geometry and update iteratively. Even though solving non-convex functions is NP-hard in the worst case, the optimization quality in practice is often not an issue -- optimizers are largely believed to find approximate global minima. Researchers hypothesize a unified explanation for this intriguing phenomenon: most of the local minima of the practically-used objectives are approximately global minima. We rigorously formalize it for concrete instances of machine learning problems.
Creating presentation materials requires complex multimodal reasoning skills to summarize key concepts and arrange them in a logical and visually pleasing manner. Can machines learn to emulate this laborious process? We present a novel task and approach for document-to-slide generation. Solving this involves document summarization, image and text retrieval, slide structure and layout prediction to arrange key elements in a form suitable for presentation. We propose a hierarchical sequence-to-sequence approach to tackle our task in an end-to-end manner. Our approach exploits the inherent structures within documents and slides and incorporates paraphrasing and layout prediction modules to generate slides. To help accelerate research in this domain, we release a dataset about 6K paired documents and slide decks used in our experiments. We show that our approach outperforms strong baselines and produces slides with rich content and aligned imagery.
Compared with cheap addition operation, multiplication operation is of much higher computation complexity. The widely-used convolutions in deep neural networks are exactly cross-correlation to measure the similarity between input feature and convolution filters, which involves massive multiplications between float values. In this paper, we present adder networks (AdderNets) to trade these massive multiplications in deep neural networks, especially convolutional neural networks (CNNs), for much cheaper additions to reduce computation costs. In AdderNets, we take the $\ell_1$-norm distance between filters and input feature as the output response. The influence of this new similarity measure on the optimization of neural network have been thoroughly analyzed. To achieve a better performance, we develop a special back-propagation approach for AdderNets by investigating the full-precision gradient. We then propose an adaptive learning rate strategy to enhance the training procedure of AdderNets according to the magnitude of each neuron's gradient. As a result, the proposed AdderNets can achieve 74.9% Top-1 accuracy 91.7% Top-5 accuracy using ResNet-50 on the ImageNet dataset without any multiplication in convolution layer.
注冊地址： 北京市海淀區羊坊店路18號2幢3層301-191