COMA: A Compositional Misleading Attack Class on Security-RAG, and a Causal Counterfactual Defense

  • 类型:arxiv
  • 标识:2608.17960
  • 链接:http://arxiv.org/abs/2608.17960v1
  • 主分类:rag
  • 形态:position
  • 被引:0
  • 被引来源:Semantic Scholar
  • S2被引:0
  • 影响力被引:0
  • TLDR:An audit is proposed that measures the leave-one-out causal influence of each retrieved document and flags answers whose influence concentrates on low-trust documents, and proposes an audit that measures the leave-one-out causal influence of each retrieved document and flags answers whose influence concentrates on low-trust documents.
  • 副分类:risk
  • 待LLM分类:否
  • 标题中文:COMA:针对 Security-RAG 的组合式误导攻击类别及因果反事实防御。
  • TLDR中文:提出一种审计方法,度量每个检索文档的留一因果影响,并对影响集中于低可信度文档的答案进行标记;并提出一种审计方法,度量每个检索文档的留一因果影响,并对影响集中于低可信度文档的答案进行标记。
  • 来源文件
  • /inbox/tom/_candidates/2026-08-19-agent-rag-longcontext-candidates.json
  • [S2 enrich]
  • /inbox/tom/_candidates/2026-08-20-agent-rag-longcontext-candidates.json