Search

An evolving solution to address hallucination and enhance accuracy in large language models (LLMs) is Retrieval-Augmented Generation (RAG), which involves augmenting LLMs with information retrieved from an external knowledge source, such as the web. This paper profiles several RAG execution pipelines and demystifies the complex interplay between their retrieval and generation phases. We demonstrate that while exact retrieval schemes are expensive, they can reduce inference time compared to approximate retrieval variants because an exact retrieval model can send a smaller but more accurate list of documents to the generative model while maintaining the same end-to-end accuracy. This observation motivates the acceleration of the exact nearest neighbor search for RAG. In this work, we design Intelligent Knowledge Store (IKS), a type-2 CXL device that implements a scale-out near-memory acceleration architecture with a novel cache-coherent interface between the host CPU and near-memory accelerators. IKS offers 13.4-27.9x faster exact nearest neighbor search over a 512GB vector database compared with executing the search on Intel Sapphire Rapids CPUs. This higher search performance translates to 1.7-26.3x lower end-to-end inference time for representative RAG applications. IKS is inherently a memory expander; its internal DRAM can be disaggregated and used for other applications running on the server to prevent DRAM, which is the most expensive component in today's servers, from being stranded.

Modern commercial-off-the-shelf (COTS) multicore processors have advanced memory hierarchies that enhance memory-level parallelism (MLP), which is crucial for high performance. To support high MLP, shared last-level caches (LLCs) are divided into multiple banks, allowing parallel access. However, uneven distribution of cache requests from the cores, especially when requests from multiple cores are concentrated on a single bank, can result in significant contention affecting all cores that access the cache. Such cache bank contention can even be maliciously induced -- known as cache bank-aware denial-of-service (DoS) attacks -- in order to jeopardize the system's timing predictability. In this paper, we propose a per-bank bandwidth regulation approach for multi-banked shared LLC based multicore real-time systems. By regulating bandwidth on a per-bank basis, the approach aims to prevent unnecessary throttling of cache accesses to non-contended banks, thus improving overall performance (throughput) without compromising isolation benefits of throttling. We implement our approach on a RISC-V system-on-chip (SoC) platform using FireSim and evaluate extensively using both synthetic and real-world workloads. Our evaluation results show that the proposed per-bank regulation approach effectively protects real-time tasks from co-running cache bank-aware DoS attacks, and offers up to a 3.66$\times$ performance improvement for the throttled benign best-effort tasks compared to prior bank-oblivious bandwidth throttling approaches., Comment: RTSS 2024

End-to-end web testing is challenging due to the need to explore diverse web application functionalities. Current state-of-the-art methods, such as WebCanvas, are not designed for broad functionality exploration; they rely on specific, detailed task descriptions, limiting their adaptability in dynamic web environments. We introduce NaviQAte, which frames web application exploration as a question-and-answer task, generating action sequences for functionalities without requiring detailed parameters. Our three-phase approach utilizes advanced large language models like GPT-4o for complex decision-making and cost-effective models, such as GPT-4o mini, for simpler tasks. NaviQAte focuses on functionality-guided web application navigation, integrating multi-modal inputs such as text and images to enhance contextual understanding. Evaluations on the Mind2Web-Live and Mind2Web-Live-Abstracted datasets show that NaviQAte achieves a 44.23% success rate in user task navigation and a 38.46% success rate in functionality navigation, representing a 15% and 33% improvement over WebCanvas. These results underscore the effectiveness of our approach in advancing automated web application testing.

Dockerfile flakiness, characterized by inconsistent build behavior without Dockerfile or project source code changes, poses significant challenges in Continuous Integration and Delivery (CI/CD) pipelines. This issue can lead to unreliable deployments and increased debugging efforts, yet it remains underexplored in current research. We conduct a systematic analysis of Dockerfile flakiness, presenting a comprehensive taxonomy of common flakiness categories, including dependency-related errors and server connectivity issues. Furthermore, we introduce FlakiDock, a tool leveraging large language models and retrieval-augmented generation techniques with dynamic analysis and an iterative feedback loop to automatically repair flaky Dockerfiles. Our evaluation shows that FlakiDock achieves a 73.55% repair accuracy, outperforming existing tools such as PARFUM by 12,581% and GPT-4-based prompting by 94.63%. These results underscore the effectiveness of FlakiDock in addressing Dockerfile flakiness and improving build reliability.

End-to-end (E2E) testing is essential for ensuring web application quality. However, manual test creation is time-consuming, and current test generation techniques produce incoherent tests. In this paper, we present AutoE2E, a novel approach that leverages Large Language Models (LLMs) to automate the generation of semantically meaningful feature-driven E2E test cases for web applications. AutoE2E intelligently infers potential features within a web application and translates them into executable test scenarios. Furthermore, we address a critical gap in the research community by introducing E2EBench, a new benchmark for automatically assessing the feature coverage of E2E test suites. Our evaluation on E2EBench demonstrates that AutoE2E achieves an average feature coverage of 79%, outperforming the best baseline by 558%, highlighting its effectiveness in generating high-quality, comprehensive test cases.

Large language models have made substantial progress in addressing diverse code-related tasks. However, their adoption is hindered by inconsistencies in generating output due to the lack of real-world, domain-specific information, such as for intra-repository API calls for unseen software projects. We introduce a novel technique to mitigate hallucinations by leveraging global and local contextual information within a code repository for API completion tasks. Our approach is tailored to refine code completion tasks, with a focus on optimizing local API completions. We examine relevant import statements during API completion to derive insights into local APIs, drawing from their method signatures. For API token completion, we analyze the inline variables and correlate them with the appropriate imported modules, thereby allowing our approach to rank the most contextually relevant suggestions from the available local APIs. Further, for conversational API completion, we gather APIs that are most relevant to the developer query with a retrieval-based search across the project. We employ our tool, LANCE, within the framework of our proposed benchmark, APIEval, encompassing two different programming languages. Our evaluation yields an average accuracy of 82.6% for API token completion and 76.9% for conversational API completion tasks. On average, LANCE surpasses Copilot by 143% and 142% for API token completion and conversational API completion, respectively. The implications of our findings are substantial for developers, suggesting that our lightweight context analysis can be applied to multilingual environments without language-specific training or fine-tuning, allowing for efficient implementation with minimal examples and effort.

The MRS mode of the JWST-MIRI instrument has been shown to be a powerful tool to characterise the molecular gas emission of the inner region of planet-forming disks. Here, we analyse the spectrum of the compact T-Tauri disk DR Tau, which is complemented by high spectral resolution (R~60000-90000) CO ro-vibrational observations. Various molecular species, including CO, CO$_2$, HCN, and C$_2$H$_2$ are detected in the JWST-MIRI spectrum, for which excitation temperatures of T~325-900 K are retrieved using LTE slab models. The high-resolution CO observations allow for a full treatment of the line profiles, which show evidence for two components of the main isotopologue, $^{12}$CO: a broad component tracing the Keplerian disk and a narrow component tracing a slow disk wind. Rotational diagrams yield excitation temperatures of T>725 K for CO, with consistently lower temperatures found for the narrow components, suggesting that the disk wind is launched from a larger distance. The inferred excitation temperatures for all molecules suggest that CO originates from the highest atmospheric layers close to the host star, followed by HCN and C$_2$H$_2$, which emit, together with $^{13}$CO, from slightly deeper layers, whereas the CO$_2$ originates from even deeper inside or further out in the disk. Additional analysis of the $^{12}$CO line wings hint at a misalignment between the inner (i~20 degrees) and outer disk (i~5 degrees). Finally, we emphasise the need for complementary high-resolution CO observations, as in combination with the JWST-MIRI observations they can be used to characterise the CO kinematics and the physical and chemical conditions of the other observed molecules with respect to CO., Comment: Accepted for publication in Astronomy & Astrophysics on 20/03/2024

Automated test generation for web forms has been a longstanding challenge, exacerbated by the intrinsic human-centric design of forms and their complex, device-agnostic structures. We introduce an innovative approach, called FormNexus, for automated web form test generation, which emphasizes deriving semantic insights from individual form elements and relations among them, utilizing textual content, DOM tree structures, and visual proximity. The insights gathered are transformed into a new conceptual graph, the Form Entity Relation Graph (FERG), which offers machine-friendly semantic information extraction. Leveraging LLMs, FormNexus adopts a feedback-driven mechanism for generating and refining input constraints based on real-time form submission responses. The culmination of this approach is a robust set of test cases, each produced by methodically invalidating constraints, ensuring comprehensive testing scenarios for web forms. This work bridges the existing gap in automated web form testing by intertwining the capabilities of LLMs with advanced semantic inference methods. Our evaluation demonstrates that FormNexus combined with GPT-4 achieves 89% coverage in form submission states. This outcome significantly outstrips the performance of the best baseline model by a margin of 25%.

Our paper addresses characterizing conditions for a linear quadratic (LQ) game to be a potential game. The desired properties of potential games in finite action settings, such as convergence of learning dynamics to Nash equilibria, and the challenges of learning Nash equilibria in continuous state and action settings motivate us to characterize LQ potential games. Our first contribution is to show that the set of LQ games with full-state feedback that are potential games is very limited, essentially differing only slightly from an identical interest game. Given this finding, we restrict the class of LQ games to those with decoupled dynamics and decoupled state information structure. For this subclass, we show that the set of potential games strictly includes non-identical interest games and characterize conditions for the LQ games in this subclass to be potential. We further derive their corresponding potential function and prove the existence of a Nash equilibrium. Meanwhile, we highlight the challenges in the characterization and computation of Nash equilibrium for this class of potential LQ games., Comment: 12 pages, 0 figure

While (1) serverless computing is emerging as a popular form of cloud execution, datacenters are going through major changes: (2) storage dissaggregation in the system infrastructure level and (3) integration of domain-specific accelerators in the hardware level. Each of these three trends individually provide significant benefits; however, when combined the benefits diminish. Specifically, the paper makes the key observation that for serverless functions, the overhead of accessing dissaggregated persistent storage overshadows the gains from accelerators. Therefore, to benefit from all these trends in conjunction, we propose Domain-Specific Computational Storage for Serverless (DSCS-Serverless). This idea contributes a serverless model that leverages a programmable accelerator within computational storage to conjugate the benefits of acceleration and storage disaggregation simultaneously. Our results with eight applications shows that integrating a comparatively small accelerator within the storage (DSCS-Serverless) that fits within its power constrains (15 Watts), significantly outperforms a traditional disaggregated system that utilizes the NVIDIA RTX 2080 Ti GPU (250 Watts). Further, the work highlights that disaggregation, serverless model, and the limited power budget for computation in storage require a different design than the conventional practices of integrating microprocessors and FPGAs. This insight is in contrast with current practices of designing computational storage that are yet to address the challenges associated with the shifts in datacenters. In comparison with two such conventional designs that either use quad-core ARM A57 or a Xilinx FPGA, DSCS-Serverless provides 3.7x and 1.7x end-to-end application speedup, 4.3x and 1.9x energy reduction, and 3.2x and 2.3x higher cost efficiency, respectively.

Enabling dexterous manipulation and safe human-robot interaction, soft robots are widely used in numerous surgical applications. One of the complications associated with using soft robots in surgical applications is reconstructing their shape and the external force exerted on them. Several sensor-based and model-based approaches have been proposed to address the issue. In this paper, a shape sensing technique based on Electrical Impedance Tomography (EIT) is proposed. The performance of this sensing technique in predicting the tip position and contact force of a soft bending actuator is highlighted by conducting a series of empirical tests. The predictions were performed based on a data-driven approach using a Long Short-Term Memory (LSTM) recurrent neural network. The tip position predictions indicate the importance of using EIT data along with pressure inputs. Changing the number of EIT channels, we evaluated the effect of the number of EIT inputs on the accuracy of the predictions. The least RMSE values for the tip position are 3.6 and 4.6 mm in Y and Z coordinates, respectively, which are 7.36% and 6.07% of the actuator's total range of motion. Contact force predictions were conducted in three different bending angles and by varying the number of EIT channels. The results of the predictions illustrated that increasing the number of channels contributes to higher accuracy of the force estimation. The mean errors of using 8 channels are 7.69%, 2.13%, and 2.96% of the total force range in three different bending angles., Comment: Revised version for 6th IEEE-RAS International Conference on Soft Robotics (RoboSoft 2023) - updated figures and discussion

Full-system simulation of computer systems is critical to capture the complex interplay between various hardware and software components in future systems. Modeling the network subsystem is indispensable to the fidelity of the full-system simulation due to the increasing importance of scale-out systems. The network software stack has undergone major changes over the last decade, and kernel-bypass networking stacks and data-plane networks are rapidly replacing the conventional kernel network stack. Nevertheless, the current state-of-the-art architectural simulators still use kernel networking which precludes realistic network application scenarios. In this work, we enable kernel bypass networking stack on gem5, the state-of-the-art full-system architectural simulator. We extend gem5's NIC hardware model and device driver to enable the support for userspace device drivers to run the DPDK framework. We also implement a network load generator hardware model in gem5 to generate various traffic patterns and perform per-packet timestamp and latency measurements without introducing packet loss. Our experimental results show that DPDK's simulated network bandwidth scales with the number of cores and NIC ports. As two use cases, we analyze the sensitivity of (1) network performance to several microarchitectural parameters, and (2) direct cache access (DCA) technology to DPDK burst size., Comment: 4 Pages, 4 Figures and 1 Table