Automated Self-Testing as a Quality Gate: Evidence-Driven Release Management for LLM Applications

2026-03-18T08:51:52Z457528060cf115d6c6bdf1af259272309f24684f29192e0ddf529a0eccf2fc5d
LLM applicationsLoosely-Structured SoftwareSATDSEMAGVIBEPASSVibeContractautomated testingbackbone algorithmscode generationcontract-driven QAdatasets mapengineering design datasetsevidence coverageevidence-based gatesfault-triggering test generationlatency monitoringmulti-agent systemsrelease governanceruntime evolutionsafety testingscientific software technical debt','code review','human-AI codeself-evolutionary agentssoftware quality assurancevariability modelsvibe coding

What happened

Collection of recent research on engineering, QA, and governance for LLM-driven and agentic software. Key contributions include: an automated self-testing quality-gate framework for LLM applications that uses evidence-based PROMOTE/HOLD/ROLLBACK decisions across five dimensions (task success, context preservation, P95 latency, safety pass rate, evidence coverage); Loosely-Structured Software (LSS) principles for managing runtime-generated entropy in multi-agent systems; VibeContract, a contract-based QA paradigm for vibe coding that decomposes intent into verifiable task-level contracts; SEMAG

Why it matters

A reviewed impact interpretation has not been published for this record.

Evidence and limitations

Source ID
arxiv_cs_se
Record identifier
457528060cf115d6c6bdf1af259272309f24684f29192e0ddf529a0eccf2fc5d
Enrichment time
2026-03-18T08:51:52Z
AI-assisted enrichment
Yes

This record may overlap with other records. Its enrichment can be incomplete or wrong, and machine assistance was used. Validate consequential decisions against the linked source and your own environment.