{
  "id": 18,
  "url": "https://arxiv.org/abs/2607.11151v1",
  "title": "AMT-X: Phase-Structured Multi-Turn Red-Teaming with Checklist-Gated Evaluation",
  "summary": "Safety evaluation of large language models (LLMs) relies largely on single-turn attack datasets and single-judge scoring, underestimating risk from adaptive multi-turn adversaries and reporting a single success rate that does not separate partially actionable outputs from those carrying complete operational detail. We propose AMT-X (Adaptive Multi-Turn Exploitation), a phase-structured multi-turn red-teaming framework. Unlike prior multi-turn attacks that rely on ad hoc escalation or free-form p",
  "authors": "Yi Ting Shen, Kentaroh Toyoda, Alex Leung",
  "category": "research",
  "topics": "safety-alignment",
  "orgs": null,
  "regions": null,
  "published_at": "2026-07-13T06:46:56.000Z",
  "fetched_at": "2026-07-14T14:14:15.663Z",
  "source_slug": "arxiv-ethics",
  "source_name": "arXiv",
  "source_homepage": "https://arxiv.org",
  "ethics_ai_record_url": "https://ethics.ai/record/18",
  "original_url": "https://arxiv.org/abs/2607.11151v1",
  "evidence_status": "source-only",
  "attribution": "via ethics.ai"
}