风险评测 · 评测基准

TamperBench: Systematically Stress-Testing LLM Safety Under Fine-Tuning and Tampering

NUWA 研究解读 · 独立复核待完成

首次发表: 2026-02-06来源核验: 2026-10-25

原始来源 ↗引用本记录 ↓

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：抗篡改是受效用约束的安全稳健性。

研究问题

开放权重防护对保留效用的修改有多稳健？

研究设计与假设

21个模型面对九类篡改，并分别搜索超参数，同时评价安全与效用。

TamperBench — Figure 8 — Figure 8 · 原文 Figure 8。请结合下方的协议范围与局限阅读。原图与来源 ↗

来源发现

主要比较在MMLU-Pro相对退化不超过10%时选择最坏StrongREJECT结果。

每模型攻击40次搜索定义了对抗预算。

NUWA 解读

NUWA 解读：抗篡改是受效用约束的安全稳健性。

局限与相反证据

主要模型为0.6–8B，聚焦拒绝型防护，大模型覆盖初步（§5）。

研究与决策意义

模型建议：比较时保留效用约束与攻击搜索预算。

精确来源位置

TamperBench : Systematically Stress-Testing LLM Safety Under Fine-Tuning and Tampering · AbstractSelected original resource https://arxiv.org/html/2602.06911; sha256:791d59fea870ef2288ea6f4c60dd83b68233441d4eeafef2b3aabc449a40b22a; archived source capture; exact capture time not independently established
TamperBench : Systematically Stress-Testing LLM Safety Under Fine-Tuning and Tampering · §4 Experiments: scoring and 40-trial searchSelected original resource https://arxiv.org/html/2602.06911; sha256:791d59fea870ef2288ea6f4c60dd83b68233441d4eeafef2b3aabc449a40b22a; archived source capture; exact capture time not independently established
TamperBench : Systematically Stress-Testing LLM Safety Under Fine-Tuning and Tampering · §5 LimitationsSelected original resource https://arxiv.org/html/2602.06911; sha256:791d59fea870ef2288ea6f4c60dd83b68233441d4eeafef2b3aabc449a40b22a; 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

研究机构滑铁卢大学 FAR.AI 多伦多大学苏黎世联邦理工学院麻省理工学院 Vector Institute

作者 (11)

Saad Hossain · Tom Tseng · Punya Syon Pandey · Samanvay Vajpayee · Matthew Kowal · Nayeema Nonta · Samuel Simko · Stephen Casper · Zhijing Jin · Kellin Pelrine · Sirisha Rambhatla

评测协议

TamperBench →

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{tamperbench,
  title = {TamperBench: Systematically Stress-Testing LLM Safety Under Fine-Tuning and Tampering},
  url = {https://arxiv.org/abs/2602.06911},
  author = {Saad Hossain and Tom Tseng and Punya Syon Pandey and Samanvay Vajpayee and Matthew Kowal and Nayeema Nonta and Samuel Simko and Stephen Casper and Zhijing Jin and Kellin Pelrine and Sirisha Rambhatla},
  year = {2026},
  date = {2026-02-06},
  version = {arXiv:2602.06911v2},
  note = {arXiv:2602.06911v2}
}

NUWA 记录 · BibTeX

@misc{tamperbench-nuwa,
  title = {TamperBench: Systematically Stress-Testing LLM Safety Under Fine-Tuning and Tampering — NUWA research record},
  url = {https://whitzard.tech/nuwa/whitzard-index/work/tamperbench/},
  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--tamperbench.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--tamperbench.json.}
}

证据快照 2026-10-29.1 · 索引更新 · RSS

参考文献

[1] Saad Hossain, Tom Tseng 等. TamperBench: Systematically Stress-Testing LLM Safety Under Fine-Tuning and Tampering. 2026-02-06. v2 [arXiv]; Author affiliations on the original paper title page; Abstract; §4 Experiments: scoring and 40-trial search