Integrating Large Language Models (LLMs) into software engineering unlocks new opportunities to automate manual processes but raises challenges around reliability, safety, and scalability. My research centers on this synergy, with two key objectives: first, harnessing LLMs to solve software engineering tasks traditionally dependent on labor-intensive, domain-specific methods, and second, applying robust software engineering principles to improve LLM safety and performance. This dual focus creates a powerful feedback loop, where LLMs drive innovation while engineering rigor ensures these systems meet the high standards required for real-world applications.

Efficiently and accurately retrieving specific information from healthcare datasets, such as the Vaccine Adverse Event Reporting System (VAERS)1, presents significant challenges. A promising solution to this problem is the Text-to-ESQ approach, which is akin to Text-to-SQL tasks but leverages NoSQL database Elasticsearch, to thoroughly explore VAERS data. Non-relational databases are particularly adept at managing complex and dynamic data formats, thereby enabling the extraction of more valuable insights. However, generating executable NoSQL queries is still challenging due to the limited availability of NoSQL query datasets, which constrains model training. One potential remedy involves the use of large language models (LLMs), which can be applied in few-shot and even zero-shot learning scenarios. Nonetheless, the lack of prior evaluation for this novel task, coupled with the absence of a comprehensive, unbiased assessment of existing LLMs and prompting strategies, impedes the development of a robust architecture. Motivated by these challenges, we introduce a new Instruction-Enhanced Explainable (InstructEx) Chain-of-Thought (CoT) prompting by integrating existing CoT prompts and conducting a comprehensive investigation of LLMs and CoT prompting. The extensive experimental analysis demonstrates the effectiveness of using LLMs for Text-to-ESQ when combined with the InstructExCoT prompting. It also sheds light on the strengths and weaknesses of these methods from multiple perspectives.

Indexing structures are widely used in modern data-processing applications to support high-performance queries, and there are a variety of recent designs specifically optimized for the newly available persistent memory (PM). The primary focus of previous PM indexes is on reducing the expensive PM writes for persisting data. However, we find that in tree-based PM indexes, because of the smaller performance gap between writes and random reads on real PM devices, the read-intensive tree traversal phase dominates the overall latency. This observation calls for further optimizations on existing indexing structures for PM. In this paper, we propose Extendible Radix Tree (ERT), an efficient indexing structure for PM that significantly reduces tree heights to minimize random reads, while still maintaining fast in-node search speed. The key idea is to use extendible hashing for each node in a radix tree. This design allows us to have a relatively large fanout of the radix tree to keep the tree height small, and also to realize constant-time lookups within a node. Using extendible hashing also allows for incremental node modification without excessive writes during inserts and updates. Range queries are efficiently and robustly handled by enforcing partial ordering among the keys in the hash table of each node without introducing more hash collisions. Our experiments on both synthetic and real-world data sets demonstrate that ERT achieves up to 2.65×, 4.41×, and 2.43× speedups for search, insert, and range queries over the respectively state-of-the-art PM index.

The recently proposed capability-based NLP testing allows model developers to test the functional capabilities of NLP models, revealing functional failures for models with good held-out evaluation scores. However, existing work on capability-based testing requires the developer to compose each individual test template from scratch. Such approach thus requires extensive manual efforts and is less scalable. In this paper, we investigate a different approach that requires the developer to only annotate a few test templates, while leveraging the GPT-3 engine to generate the majority of test cases. While our approach saves the manual efforts by design, it guarantees the correctness of the generated suites with a validity checker. Moreover, our experimental results show that the test suites generated by GPT-3 are more diverse than the manually created ones; they can also be used to detect more errors compared to manually created counterparts. Our test suites can be downloaded at https://anonymous-researcher-nlp.github.io/testaug/.

Deploying wireless sensor networks to provide guaranteed barrier coverage is critical for many sensor networks applications such as intrusion detection and border surveillance. To reduce the number of sensors needed to provide guaranteed barrier coverage, we propose multi-round sensor deployment which splits sensor deployment into multiple rounds and can better deal with placement errors that often accompany sensor deployment. We conduct a comprehensive analytical study on multi-round sensor deployment and identify the tradeoff between the number of sensors deployed in each round of multi-round sensor deployment and the barrier coverage performance. Both numerical and simulation studies show that, by simply splitting sensor deployment into two rounds, guaranteed barrier coverage can be achieved with significantly less sensors comparing to single-round sensor deployment. Moreover, we propose two practical solutions for multi-round sensor deployment when the distribution of a sensor's residence point is not fully known. The effectiveness of the proposed multi-round sensor deployment strategies is demonstrated by numerical and simulation results.

Sensor networks have been deployed for many barrier coverage applications such as intrusion detection and border surveillance. In these applications, it is critical to operate a sensor network in an energy-efficient manner so the barrier can be covered with as few active sensors as possible. In this paper, we study barrier information coverage which exploits collaborations and information fusion between neighboring sensors to reduce the number of active sensors needed to cover a barrier and hence to prolong the network lifetime. Moreover, we propose a practical solution to identify the barrier information coverage set which can information-cover the barrier with a small number of active sensors. The effectiveness of the proposed solution is demonstrated by numerical and simulation results.

Quality of object detection and network lifetime hold critical importance to many sensor network applications such as military surveillance. Unfortunately, improving one of these aspects comes at the expense of the other. In this paper, based on the probabilistic sensing model, we propose a novel framework for object detection in sensor networks, called DeCODe (on-demand framework for collaborative object detection), which provides a desired object detection performance (characterized in terms of detection probability and false detection probability), while attempting to prolong the network lifetime. The design of DeCODe is motivated by a counterintuitive observation that simple collaboration among active sensors indeed degrades the object detection performance. By contrast, each active sensor in DeCODe can trigger its neighboring inactive sensors to participate in the detection process in an on-demand fashion, so as to achieve the same low false detection probability while increasing the probability of detection. The effectiveness of the proposed DeCODe framework is supported by theoretical analysis and simulation-based validation.

The overlay-based network architecture has been recognized as an effective way to deal with the funneling effect in sensor networks, where sensors closer to the sink are usually responsible for relaying more network traffic. Such funneling effect is particularly harmful when the number of sensors in the network is vast. In an overlay-based sensor network, a special type of resource-rich multi-radio mobile wireless devices (we call them syphons) are deployed along with sensors. Syphons form an overlay network and help nearby sensors relay their data to the sink via the overlay network, thus mitigating the funneling effect. In this paper, we study one of the fundamental challenges in overlay-based sensor networks: syphon deployment problem, i.e., how to deploy a limited number of syphons to cover a vast sensing field while maintaining the connectivity and balanced loads among them. We propose a novel sensor-aided overlay deployment and relocation (SODaR) protocolas a possible solution. The key idea is to take advantage of sensors' assistance and to relocate syphons by circling them around the sink in an orderly manner until all syphons are connected. Simulation results show that, with SODaR, syphons are able to self-form and self-maintain a connected tree structure which provides excellent load balancing among syphons with modest message and movement overhead.

Many sensor networks are deployed for the purpose of covering and monitoring a particular region, and detecting the object of interest in the region. In this paper, based on the probabilistic sensing model, we conduct comprehensive analytical and simulation studies on collaborative information coverage and object detection in wireless sensor networks. More specifically, we first define point information coverage and based on that we define p-coverage as a measure of the coverage performance for a randomly-deployed wireless sensor network. Then, we investigate the coverage and object detection performances of several simple decision fusion-based collaborative mechanisms and find that simple collaborations among active sensors indeed degrade the coverage performance due to the requirement of maintaining the target false detection probability. This motivates us to develop an on-demand collaborative framework for object detection, whose effectiveness is supported by detailed theoretical analysis and simulation-based validation. Finally, we investigate the energy efficiency performance of our proposed framework and identify the trade-offs among various system parameters in network power consumption.