研究定位 LLaMA-2-7B-Chat 安全敏感参数
定位安全敏感参数并稀疏改动权重以削弱安全对齐
- arXiv
- 2610.09000
- Published
- Layer
- Model layer
- Setting
- Models and APIs
- Evaluated models
- LLaMA-2-7B-Chat
- AttacksModel tampering
- RiskInappropriate refusal
All papers
Filter5
Attack type
Impact
System type
Layer
Select any value within a category; different categories must all match. Counts include the other filters you have selected.Another 215papers have no tags yet and are not included in filtered results.
Another 215papers do not have category tags yet and are not included in filtered results.
TRACE：仅凭 checkpoint 更新审计大语言模型有害目标
提出 TRACE，仅凭检查点更新审计有害合规微调目标
- arXiv
- 2610.07518
- Published
- Setting
- Models and APIs
- Evaluated models
- Llama-3.1-8B-Instruct, Mistral-7B-Instruct-v0.3, Qwen3-8B and others6
- AttacksModel tampering
- ComponentFine-tuning and open weights
Abstract作者称有害合规 SFT 在检查点更新中留下可排序坐标，并据此做只读权重审计。
作者在检查点更新里寻找有害合规 SFT 的目标组成，并把读数做成不查询模型、不看未知训练数据的审计分 TRACE。
用权重谱密度预测隐私泄露：WeightWatcher 谱指标可作为成员推理攻击脆弱性的代理
检验权重谱度量能否代理成员推断泄漏
- arXiv
- 2609.11780
- Published
- Layer
- Model layer
- Setting
- Models and APIs
- Evaluated models
- MLP-1×4096, MLP-2×1024, MLP-2×2048 and others11
- AttacksExtraction and theft
Abstract作者认为谱度量与泛化差距互补，有望用于无数据的模型级筛查。
作者在 MLP 上检查 WeightWatcher 谱统计是否与 online LiRA 同向，以便不看数据、不训练 shadow 模型就做模型级成员风险筛查。放出前的审计要知道训练成员会不会被认出来。LiRA 依赖大量 shadow 模型，候选一多就难以逐个跑。
论文提出拒答的几何解释：事后安全训练脆弱，预训练期持续安全共训更抗攻击
解释后验安全的权重几何脆弱性，并提出预训练持续安全共训练
- arXiv
- 2609.06934
- Published
- Layer
- Model layer
- Setting
- Models and APIs
- Evaluated models
- Qwen-2.5-7B, Qwen-2.5-7B-Instruct, Llama-3-8B and others15
- DefensesModel hardening
- AttacksModel tampering
- RiskInappropriate refusal
- ComponentFine-tuning and open weights
Abstract论文将后验安全脆弱性归因于正交抑制，并提出通过预训练持续共训练实现具备微调抗性的安全防护。
You have reached the end