The recent developments in the field of software testing and quality assurance, particularly in the context of Deep Learning (DL) systems and Large Language Models (LLMs), indicate a significant shift towards more efficient, stable, and context-aware testing methodologies. Innovations are focusing on enhancing the stability and sensitivity of mutation testing for DL systems, improving the granularity and coverage of unit testing for autonomous driving software, and leveraging LLMs for automated refactoring engine testing and bug localization. Additionally, there's a growing interest in comparing traditional test generation methods with LLM-based approaches, exploring the potential of LLMs in regression test generation, and addressing the challenges of flaky tests in JavaScript. The field is also seeing advancements in accelerating deep neural network mutation analysis and developing adaptive testing strategies for LLM-based applications.

Noteworthy papers include:

• MuFF: Introduces a post-training DL mutation technique ensuring mutant stability and sensitivity, significantly outperforming existing methods.
• Fine-grained Testing for Autonomous Driving Software: Presents a novel approach to enhance test coverage and pass rates in Autoware unit testing using LLMs.
• Testing Refactoring Engine via Historical Bug Report driven LLM: Proposes RETESTER, a framework that successfully identified new bugs in popular refactoring engines.
• Test Wars: Compares SBST, symbolic execution, and LLM-based test generation, highlighting the semantic understanding strengths of LLMs.
• Improved IR-based Bug Localization with Intelligent Relevance Feedback: Introduces BRaIn, significantly outperforming existing bug localization techniques.
• Can LLM Generate Regression Tests for Software Commits?: Explores LLM-based regression test generation, showing promise for human-readable input formats.
• On Accelerating Deep Neural Network Mutation Analysis by Neuron and Mutant Clustering: Presents DEEPMAACC, a technique that significantly speeds up DNN mutation analysis.
• Detecting and Evaluating Order-Dependent Flaky Tests in JavaScript: Investigates test order dependency in JavaScript, identifying shared mocking state as a new cause of flakiness.
• Mutation-Guided LLM-based Test Generation at Meta: Describes Meta's ACH system for generating privacy-hardening test cases using LLMs.
• Adaptive Testing for LLM-Based Applications: Proposes a diversity-based approach to testing LLM-based applications, reducing testing budgets and generating varied outputs.