This paper introduces Requirements-Augmented Generation (REAG), which interprets user intentions by retrieving relevant software requirements, domain knowledge, and personas via adaptive RAG and self-reasoning to generate context-aware test oracles, and introduces a confidence-calibrated cascade judgment, which quantifies verdict reliability via simulated expert agreement.
Abstract
LLM-based software (LBS) integrates large language models as core components to deliver flexible, personalised responses. Unlike traditional software with deterministic outputs, LBSs exhibit context-dependent, stochastic behaviour that renders classical acceptance testing and test oracles insufficient: the same query may require fundamentally different responses depending on user personas and software context. This gap creates an urgent need for automated acceptance testing frameworks that autonomously interpret user instructions, while reliably inferring user intentions in a changing environment. In this paper, we present an automated acceptance testing framework for LBS with calibrated verdict reliability via two technical contributions. First, we introduce Requirements-Augmented Generation (REAG), which interprets user intentions by retrieving relevant software requirements, domain knowledge, and personas via adaptive RAG and self-reasoning to generate context-aware test oracles. Second, recognising that oracle generation may retrieve irrelevant constraints, misinterpret intent, or hallucinate requirements, we introduce a confidence-calibrated cascade judgment. This method quantifies verdict reliability via simulated expert agreement -- accepting high-confidence verdicts, escalating ambiguous cases, or abstaining when uncertain -- with empirical reliability guarantees backed by conformal risk control. An industrial case study on a production nutrition advisory application demonstrates that REAG achieves a 3.91/5 oracle quality score, reaching qualified or marginal oracle quality in 82% of cases. The confidence-calibrated cascade achieves 98.8% accuracy, improves oracle quality from 3.91 to 4.30 by filtering unqualified outputs, and delivers a 31.7% cost-efficiency improvement over single-judge baselines, validating industrial viability
NL2Test is presented, an end-to-end approach and tool that generates executable API regression tests from a natural-language scenario description and a traffic capture recorded while executing the scenario, indicating that traffic-grounded generation with deterministic guardrails can substantially reduce manual effort while improving regression automation in complex microservice environments.
Hao-Zhen You, Zhen Dong, Jing-Jing Wang et al.· arXiv.org· 0 citations
Metamorphic Testing (MT) provides a promising approach for testing software without defined test oracles by specifying expected relations between inputs and outputs, instead of relying on exact outputs. For example, testing Augmented Reality (AR) applications is challenging due to dynamic interactions between virtual content, physical environments, and code, which make traditional test oracles difficult to define. However, formulating metamorphic relations (MRs) is time-consuming and burdensome. We introduce a context-aware pipeline that generates and refines MRs using repository-level context and reasoning orchestration, evaluated on a dataset of 142 mobile AR system repositories. Across three context configurations generating 14,916 candidate MRs, hierarchical context yielded the broadest coverage (7,004 MRs across 142 repositories and 5,167 class--method pairs) and lower redundancy. An agentic deliberation process then reconciled conflicting candidates---observed in 79.0% of cases---reducing duplication and selecting context-aware relations in 88.2% of outcomes. A manual oracle study shows refined relations (n = 141) are both logically valid and sufficiently concrete to be directly translated into test assertions, and a preliminary case study reveals converting generated MRs (n = 5) into executable tests can detect non-equivalent mutations in real-world code. Overall, our results show that combining repository-aware MR generation with reasoning-based refinement enables scalable construction of reliable, domain-relevant test oracles.
Dibyendu Brinto Bose, Jiawei Qin, Chris Brown· arXiv.org· 0 citations
AssertMate is proposed, a novel agent-based assertion generation framework that enhances the quality and reliability of LLM-generated assertions through three key components: actual value construction that identifies assertion targets via static analysis and type-aware heuristics, and multi-perspective expected value prediction using code generation, retrieval-augmented generation (RAG), and chain-of-thought (CoT) reasoning agents.
Dong Wang, Qiaoyu Han, Lin Yang et al.· 0 citations
Results support a focused conclusion: LLM-generated review is most useful as complementary semantic guidance when paired with deployment-oriented test selection, rather than as a standalone testing artifact.
Hui-Xiang Zhen, Zhihan Zhang· International Conference on...· 0 citations
Jailbreak robustness has become central to large language model (LLM) safety evaluation, yet prevailing methodologies rely primarily on refusal behavior, semantic resemblance, and intent-matching heuristics that emphasize linguistic plausibility rather than correctness. We identify a key limitation in existing evaluations: many jailbreak intents depend on instructional validity rather than epistemic factuality, allowing realistic-looking responses to be labeled successful despite being factually or procedurally incorrect. To address this gap, we propose Sequential Epistemic and Action-Level Validation (SEAV), a verification-centric jailbreak evaluation framework that decomposes responses into ordered steps and evaluates both validity and correctness. SEAV combines LLM-as-a-judge mechanisms for semantic interpretation with retrieval-grounded verification using external knowledge sources, assessing whether generated content is factually correct, structurally consistent, and operationally capable of advancing harmful objectives. Empirically, SEAV cuts the false-positive rate on SD-A (a curated strategic-dishonesty diagnostic) by 14.9\,pp vs. the strongest baseline, and reclassifies 22.1\%--51.0\% of sampled prior-labeled successes as invalid across three of four public benchmarks. Together, these results show that enforcing correctness substantially reshapes measured robustness: many previously labeled jailbreak successes are reclassified as invalid, and results are stable across the tested search backends and evaluator models. Code and data are available at https://github.com/Ardor-Wu/SEAV.
Qilong Wu, Sahil Wadhwa, Pranab Mohanty et al.· 0 citations
JSTestCraft reconstructs missing context via three enrichment agents: library, structural, and semantic via three enrichment agents, demonstrating that context reconstruction and agentic collaboration enable LLMs to perform more reliable and adaptive testing in dynamic JavaScript environments.
