风险评测 · 评测基准

BioTIER: A Refusal Benchmark for Targeted Biological Risk Mitigation

NUWA 研究解读 · 独立复核待完成

首次发表: 2026-07-16来源核验: 2026-10-25

原始来源 ↗引用本记录 ↓化生放核 →

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：定向访问政策遵从不同于生物能力或实际伤害。

研究问题

生物防护能否区分狭窄高风险请求与良性科研访问？

研究设计与假设

542项专家提示分三个风险集合；对十家开发者52个模型以实质回答或拒绝作二元判分。

BioTIER — 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

评价分别检查拒绝受限信息和允许良性信息。

二元遵从不评分正确性、完整性或现实增益（§3.3）。

NUWA 解读

NUWA 解读：定向访问政策遵从不同于生物能力或实际伤害。

局限与相反证据

分类阈值含专家判断，须定期更新，且存在针对基准训练的可能（§5.4）。

研究与决策意义

模型建议：分别报告拒绝不足与过度拒绝，不将遵从视为操作性伤害分数。

精确来源位置

BioTIER: A Refusal Benchmark for Targeted Biological Risk Mitigation · AbstractSelected original resource https://arxiv.org/html/2607.14479v1; sha256:d39bab620389ee6f797ce66762dc0bda4f5afdf89d5653bcbd08a2f0970b9e01; archived source capture; exact capture time not independently established
BioTIER: A Refusal Benchmark for Targeted Biological Risk Mitigation · §3.3 EvaluationSelected original resource https://arxiv.org/html/2607.14479v1; sha256:d39bab620389ee6f797ce66762dc0bda4f5afdf89d5653bcbd08a2f0970b9e01; archived source capture; exact capture time not independently established
BioTIER: A Refusal Benchmark for Targeted Biological Risk Mitigation · §5.4 LimitationsSelected original resource https://arxiv.org/html/2607.14479v1; sha256:d39bab620389ee6f797ce66762dc0bda4f5afdf89d5653bcbd08a2f0970b9e01; 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 →

机构与角色

研究机构 SecureBio 巴西 ABC 联邦大学

作者 (9)

Eleanor M. Marshall · Pedro Medeiros · Peter Peneder · Nelly Mak · Jacob Kaffey · Faith Rovenolt · Mac Walker · Seth Donoughe · Jasper Götting

评测协议

BioTIER →

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{biotier,
  title = {BioTIER: A Refusal Benchmark for Targeted Biological Risk Mitigation},
  url = {https://arxiv.org/abs/2607.14479v1},
  author = {Eleanor M. Marshall and Pedro Medeiros and Peter Peneder and Nelly Mak and Jacob Kaffey and Faith Rovenolt and Mac Walker and Seth Donoughe and Jasper Götting},
  year = {2026},
  date = {2026-07-16},
  version = {arXiv:2607.14479v1},
  note = {arXiv:2607.14479v1}
}

NUWA 记录 · BibTeX

@misc{biotier-nuwa,
  title = {BioTIER: A Refusal Benchmark for Targeted Biological Risk Mitigation — NUWA research record},
  url = {https://whitzard.tech/nuwa/whitzard-index/work/biotier/},
  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--biotier.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--biotier.json.}
}

证据快照 2026-10-29.1 · 索引更新 · RSS

参考文献

[1] Eleanor M. Marshall, Pedro Medeiros 等. BioTIER: A Refusal Benchmark for Targeted Biological Risk Mitigation. 2026-07-16. v1 [arXiv]; Author affiliations on the original paper title page; Abstract; §3.3 Evaluation