Beyond QASecClaw: Agent Scaling, Verifier Feedback, and Evidence Reliability in Software Assurance.

arXiv:2610.36947v1 Announce Type: cross Abstract: Can additional specialist language-model contexts improve software assurance when deterministic tool evidence is already available? We study this question through five experiments spanning authored Java warning