风险评测 · 评测基准

Detecting Multi-Agent Collusion Through Multi-Agent Interpretability

NUWA 研究解读 · 独立复核待完成

首次发表: 2026-04-01来源核验: 2026-10-25

风险监测

原始来源 ↗引用本记录 ↓欺骗 →

NUWA Research Note

研究问题研究设计与假设来源发现 NUWA 解读局限与相反证据研究与决策意义机构与角色原始资源

模型辅助整理 · 选读原始全文或官方文章中的相关章节；尚未完成逐章节审计。

研究速读

NUWA 解读：排序判别高不保证低误报部署检出。

研究问题

激活群体检测能否从已知合谋泛化到新场景？

研究设计与假设

四个开放模型、六种探针架构，在Core留域验证，再以五种聚合方法测Transfer与Stego。

Detecting Multi-Agent Collusion Through Multi-Agent Interpretability — Figure 6 — Figure 6 · 原文 Figure 6。请结合下方的协议范围与局限阅读。原图与来源 ↗

来源发现

各模型分布内AUROC均1.00。

结构迁移上最佳方法在1%误报时召回仅0.38，尽管AUROC较高。

NUWA 解读

NUWA 解读：排序判别高不保证低误报部署检出。

局限与相反证据

合成合谋与少量模型限制泛化，附录使用OOD均值选超参数，须明确调参可见性。

研究与决策意义

模型建议：报告严格阈值召回，并分开调参、训练与保留环境。

精确来源位置

Detecting Multi-Agent Collusion Through Multi-Agent Interpretability · AbstractSelected original resource https://arxiv.org/html/2604.01151v2; sha256:4a8eeea1f124ad8e96c604e36d0ecf2429f2dd528655f3aa19dde2aad705adc7; archived source capture; exact capture time not independently established
Detecting Multi-Agent Collusion Through Multi-Agent Interpretability · §3.2 Activation extraction and probe training: EvaluationSelected original resource https://arxiv.org/html/2604.01151v2; sha256:4a8eeea1f124ad8e96c604e36d0ecf2429f2dd528655f3aa19dde2aad705adc7; archived source capture; exact capture time not independently established
Detecting Multi-Agent Collusion Through Multi-Agent Interpretability · §4.4 Detection at operational thresholdsSelected original resource https://arxiv.org/html/2604.01151v2; sha256:4a8eeea1f124ad8e96c604e36d0ecf2429f2dd528655f3aa19dde2aad705adc7; archived source capture; exact capture time not independently established
Detecting Multi-Agent Collusion Through Multi-Agent Interpretability · Appendix B Logistic Regression Auxiliary ExperimentsSelected original resource https://arxiv.org/html/2604.01151v2; sha256:4a8eeea1f124ad8e96c604e36d0ecf2429f2dd528655f3aa19dde2aad705adc7; archived source capture; exact capture time not independently established

来源核查、数据提取、编辑批准与实验复现分别记录。方法说明 →

多智能体效应

NARCBench 使用四个智能体：串谋实验明确指定两个串谋者，对照实验全部诚实。探针从 Core 场景学习欺骗方向，再用五种汇总方法对整次运行分类。四个开放权重模型分别在留出 Core 领域、结构不同的 Transfer 场景和隐写纸牌游戏层级接受评估。

机构与角色

研究机构牛津大学 ELLIS Institute Tübingen

作者 (6)

Aaron Rose · Carissa Cullen · Sahar Abdelnabi · Philip Torr · Brandon Gary Kaplowitz · Christian Schroeder de Witt

评测协议

Detecting Multi-Agent Collusion Through Multi-Agent Interpretability →

引用与复用

引用研究发现时请引用原始出版物；引用 NUWA 的整理与解释时，单独引用本站记录及对应快照。

BibTeX ↓RIS ↓记录与来源元数据 · JSON ↓阅读覆盖 · JSON 提交更正

原始出版物 · BibTeX

@misc{narcbench,
  title = {Detecting Multi-Agent Collusion Through Multi-Agent Interpretability},
  url = {https://arxiv.org/abs/2604.01151v2},
  author = {Aaron Rose and Carissa Cullen and Sahar Abdelnabi and Philip Torr and Brandon Gary Kaplowitz and Christian Schroeder de Witt},
  year = {2026},
  date = {2026-04-01},
  version = {arXiv:2604.01151v2},
  note = {arXiv:2604.01151v2}
}

NUWA 记录 · BibTeX

@misc{narcbench-nuwa,
  title = {Detecting Multi-Agent Collusion Through Multi-Agent Interpretability — NUWA research record},
  url = {https://whitzard.tech/nuwa/whitzard-index/work/narcbench/},
  version = {2026-10-29.1},
  organization = {NUWA Frontier AI Safety Lab},
  howpublished = {Retained note edition: https://whitzard.tech/nuwa/whitzard-index/data/record-editions/2026-10-29.1--narcbench.json},
  note = {NUWA research note. Source-reading scope and independent review status are stated on the record. Snapshot 2026-10-29.1. Retained note edition: https://whitzard.tech/nuwa/whitzard-index/data/record-editions/2026-10-29.1--narcbench.json.}
}

证据快照 2026-10-29.1 · 索引更新 · RSS

参考文献

[1] Aaron Rose, Carissa Cullen 等. Detecting Multi-Agent Collusion Through Multi-Agent Interpretability. 2026-04-01. v2 [arXiv]; Author affiliations on the original paper title page; Abstract; §3.2 Activation extraction and probe training: Evaluation