As Graph Neural Networks (GNNs) become more pervasive, it becomes paramount to build robust tools for computing explanations of their predictions. A key desideratum is that these explanations are faithful, i.e., that they portray an accurate picture of the GNN's reasoning process. A number of different faithfulness metrics exist, begging the question of what faithfulness is exactly, and what its properties are. We begin by showing that existing metrics are not interchangeable -- i.e., explanations attaining high faithfulness according to one metric may be unfaithful according to others -- and can be systematically insensitive to important properties of the explanation, and suggest how to address these issues. We proceed to show that, surprisingly, optimizing for faithfulness is not always a sensible design goal. Specifically, we show that for injective regular GNN architectures, perfectly faithful explanations are completely uninformative. The situation is different for modular GNNs, such as self-explainable and domain-invariant architectures, where optimizing faithfulness does not compromise informativeness, and is also unexpectedly tied to out-of-distribution generalization.
相關內容
Gate-level clocking, typical in traditional approaches to Single Flux Quantum (SFQ) technology, makes the effective synthesis of superconducting circuits a significant engineering hurdle. This paper addresses this challenge by employing the recently introduced alternating SFQ (xSFQ) logic family. xSFQ leverages dual-rail alternating encoding to eliminate the clock dependency from the superconducting gate semantics. This obviates the need for ad hoc modifications to existing synthesis tools and avoids unnecessary circuit resource overheads, marking a significant advancement in superconducting circuit design automation. Our implementation results demonstrate an average reduction of over 80\% in the Josephson junction count for circuits from the ISCAS85, EPFL, and ISCAS89 benchmark suites.
Artificial intelligence (AI) hiring tools have revolutionized resume screening, and large language models (LLMs) have the potential to do the same. However, given the biases which are embedded within LLMs, it is unclear whether they can be used in this scenario without disadvantaging groups based on their protected attributes. In this work, we investigate the possibilities of using LLMs in a resume screening setting via a document retrieval framework that simulates job candidate selection. Using that framework, we then perform a resume audit study to determine whether a selection of Massive Text Embedding (MTE) models are biased in resume screening scenarios. We simulate this for nine occupations, using a collection of over 500 publicly available resumes and 500 job descriptions. We find that the MTEs are biased, significantly favoring White-associated names in 85.1\% of cases and female-associated names in only 11.1\% of cases, with a minority of cases showing no statistically significant differences. Further analyses show that Black males are disadvantaged in up to 100\% of cases, replicating real-world patterns of bias in employment settings, and validate three hypotheses of intersectionality. We also find an impact of document length as well as the corpus frequency of names in the selection of resumes. These findings have implications for widely used AI tools that are automating employment, fairness, and tech policy.
The Robot Operating System (ROS) is a popular framework and ecosystem that allows developers to build robot software systems from reusable, off-the-shelf components. Systems are often built by customizing and connecting components via configuration files. While reusable components theoretically allow rapid prototyping, ensuring proper configuration and connection is challenging, as evidenced by numerous questions on developer forums. Developers must abide to the often unchecked and unstated assumptions of individual components. Failure to do so can result in misconfigurations that are only discovered during field deployment, at which point errors may lead to unpredictable and dangerous behavior. Despite misconfigurations having been studied in the broader context of software engineering, robotics software (and ROS in particular) poses domain-specific challenges with potentially disastrous consequences. To understand and improve the reliability of ROS projects, it is critical to identify the types of misconfigurations faced by developers. To that end, we perform a study of ROS Answers, a Q&A platform, to identify and categorize misconfigurations that occur during ROS development. We then conduct a literature review to assess the coverage of these misconfigurations by existing detection techniques. In total, we find 12 high-level categories and 50 sub-categories of misconfigurations. Of these categories, 27 are not covered by existing techniques. To conclude, we discuss how to tackle those misconfigurations in future work.
In Advanced Persistent Threat (APT) attacks, achieving stealthy persistence within target systems is often crucial for an attacker's success. This persistence allows adversaries to maintain prolonged access, often evading detection mechanisms. Recognizing its pivotal role in the APT lifecycle, this paper introduces Cyber Persistence Detector (CPD), a novel system dedicated to detecting cyber persistence through provenance analytics. CPD is founded on the insight that persistent operations typically manifest in two phases: the "persistence setup" and the subsequent "persistence execution". By causally relating these phases, we enhance our ability to detect persistent threats. First, CPD discerns setups signaling an impending persistent threat and then traces processes linked to remote connections to identify persistence execution activities. A key feature of our system is the introduction of pseudo-dependency edges (pseudo-edges), which effectively connect these disjoint phases using data provenance analysis, and expert-guided edges, which enable faster tracing and reduced log size. These edges empower us to detect persistence threats accurately and efficiently. Moreover, we propose a novel alert triage algorithm that further reduces false positives associated with persistence threats. Evaluations conducted on well-known datasets demonstrate that our system reduces the average false positive rate by 93% compared to state-of-the-art methods.
In this work, a distributed server system composed of multiple servers that holds some coded files and multiple users that are interested in retrieving the linear functions of the files is investigated, where the servers are robust, blind and adversarial in the sense that any $J$ servers can together recover all files, while any $I$ colluding servers cannot obtain any information about the files, and at most $A$ servers maliciously provides erroneous information. In addition, the file library must be secure from a wiretapper who obtains all the signals, and the demands of any subset of users must kept private from the other users and servers, even if they collude. A coding scheme is proposed by incorporating the ideas of Shamir's secret sharing and key superposition into the framework of Placement Delivery Array (PDA), originally proposed to characterize the single-server coded caching system without any security or privacy constraints. It is shown that PDAs associated to Maddah-Ali and Niesen's coded caching scheme results in an achievable memory-storage-communication region, such that the storage size and communication load were optimal to within a multiplicative gap, except for the small memory regime when the number of files was smaller than the number of users.
[Context] The adoption of micro-frontends architectures has gained traction as a promising approach to enhance modularity, scalability, and maintainability of web applications. [Goal] The primary aim of this research is to investigate the benefits and limitations of migrating a real-world application to a micro-frontends architecture from the perspective of the developers. [Method] Based on the action research approach, after diagnosis and planning, we applied an intervention of migrating the target web application to a micro-frontends architecture. Thereafter, the migration was evaluated in a workshop involving the remaining developers responsible for maintaining the application. During the workshop, these developers were presented with the migrated architecture, conducted a simple maintenance task, discussed benefits and limitations in a focus group to gather insights, and answered a questionnaire on the acceptance of the technology. [Results] Developers' perceptions gathered during the focus group reinforce the benefits and limitations reported in the literature. Key benefits included enhanced flexibility in technology choices, scalability of development teams, and gradual migration of technologies. However, the increased complexity of the architecture raised concerns among developers, particularly in dependency and environment management, debugging, and integration testing. [Conclusions] While micro-frontends represent a promising technology, unresolved issues still limit their broader applicability. Developers generally perceived the architecture as useful and moderately easy to use but hesitated to adopt it.
Several open-source systems, such as Flower and NVIDIA FLARE, have been developed in recent years while focusing on different aspects of federated learning (FL). Flower is dedicated to implementing a cohesive approach to FL, analytics, and evaluation. Over time, Flower has cultivated extensive strategies and algorithms tailored for FL application development, fostering a vibrant FL community in research and industry. Conversely, FLARE has prioritized the creation of an enterprise-ready, resilient runtime environment explicitly designed for FL applications in production environments. In this paper, we describe our initial integration of both frameworks and show how they can work together to supercharge the FL ecosystem as a whole. Through the seamless integration of Flower and FLARE, applications crafted within the Flower framework can effortlessly operate within the FLARE runtime environment without necessitating any modifications. This initial integration streamlines the process, eliminating complexities and ensuring smooth interoperability between the two platforms, thus enhancing the overall efficiency and accessibility of FL applications.
Aligning future system design with the ever-increasing compute needs of large language models (LLMs) is undoubtedly an important problem in today's world. Here, we propose a general performance modeling methodology and workload analysis of distributed LLM training and inference through an analytical framework that accurately considers compute, memory sub-system, network, and various parallelization strategies (model parallel, data parallel, pipeline parallel, and sequence parallel). We validate our performance predictions with published data from literature and relevant industry vendors (e.g., NVIDIA). For distributed training, we investigate the memory footprint of LLMs for different activation re-computation methods, dissect the key factors behind the massive performance gain from A100 to B200 ($\sim$ 35x speed-up closely following NVIDIA's scaling trend), and further run a design space exploration at different technology nodes (12 nm to 1 nm) to study the impact of logic, memory, and network scaling on the performance. For inference, we analyze the compute versus memory boundedness of different operations at a matrix-multiply level for different GPU systems and further explore the impact of DRAM memory technology scaling on inference latency. Utilizing our modeling framework, we reveal the evolution of performance bottlenecks for both LLM training and inference with technology scaling, thus, providing insights to design future systems for LLM training and inference.
Remarkable progress in the development of Deep Learning Weather Prediction (DLWP) models positions them to become competitive with traditional numerical weather prediction (NWP) models. Indeed, a wide number of DLWP architectures -- based on various backbones, including U-Net, Transformer, Graph Neural Network (GNN), and Fourier Neural Operator (FNO) -- have demonstrated their potential at forecasting atmospheric states. However, due to differences in training protocols, forecast horizons, and data choices, it remains unclear which (if any) of these methods and architectures are most suitable for weather forecasting and for future model development. Here, we step back and provide a detailed empirical analysis, under controlled conditions, comparing and contrasting the most prominent DLWP models, along with their backbones. We accomplish this by predicting synthetic two-dimensional incompressible Navier-Stokes and real-world global weather dynamics. In terms of accuracy, memory consumption, and runtime, our results illustrate various tradeoffs. For example, on synthetic data, we observe favorable performance of FNO; and on the real-world WeatherBench dataset, our results demonstrate the suitability of ConvLSTM and SwinTransformer for short-to-mid-ranged forecasts. For long-ranged weather rollouts of up to 365 days, we observe superior stability and physical soundness in architectures that formulate a spherical data representation, i.e., GraphCast and Spherical FNO. In addition, we observe that all of these model backbones ``saturate,'' i.e., none of them exhibit so-called neural scaling, which highlights an important direction for future work on these and related models.
Recent innovations in artificial intelligence (AI), primarily powered by large language models (LLMs), have transformed how programmers develop and maintain software -- leading to new frontiers in software engineering (SE). The advanced capabilities of LLM-based programming assistants to support software development tasks have led to a rise in the adoption of LLMs in SE. However, little is known about the evidenced-based practices, tools and processes verified by research findings, supported and adopted by AI programming assistants. To this end, our work conducts a preliminary evaluation exploring the beliefs and behaviors of LLM used to support software development tasks. We investigate 17 evidence-based claims posited by empirical SE research across five LLM-based programming assistants. Our findings show that LLM-based programming assistants have ambiguous beliefs regarding research claims, lack credible evidence to support responses, and are incapable of adopting practices demonstrated by empirical SE research to support development tasks. Based on our results, we provide implications for practitioners adopting LLM-based programming assistants in development contexts and shed light on future research directions to enhance the reliability and trustworthiness of LLMs -- aiming to increase awareness and adoption of evidence-based SE research findings in practice.
小貼士
登錄享
相關主題
注冊地址： 北京市海淀區羊坊店路18號2幢3層301-191