# model-merging
標記為「model-merging」的 20 篇文章
Training & Fine-Tuning Attacks
Methodology for data poisoning, trojan/backdoor insertion, clean-label attacks, LoRA backdoors, sleeper agent techniques, and model merging attacks targeting the LLM training pipeline.
Model Merging Risks
Security risks in model and adapter merging workflows -- how merging adapters from untrusted sources can introduce vulnerabilities, exploit merge algorithm properties, and cause safety property loss through TIES, DARE, SLERP, and linear interpolation.
Model Merging Security Analysis
Security implications of model merging techniques (TIES, DARE, SLERP) including backdoor propagation and safety property degradation.
Model Merging Attack Surface Analysis
Security analysis of model merging techniques including TIES, DARE, and SLERP for injecting malicious capabilities.
Model Merging Security Implications
Security analysis of model merging techniques and potential for backdoor propagation through merged models.
Lab: Model Merging Security Analysis
Analyze security implications of model merging techniques and test for backdoor propagation through merged model weights.
Model Merging Backdoor Propagation
Demonstrate how backdoors propagate through model merging techniques like TIES, DARE, and spherical interpolation.
Advanced Training Attack Vectors
Cutting-edge training attacks: federated learning poisoning, model merging exploits, distributed training vulnerabilities, emergent capability risks, and synthetic data pipeline attacks.
Model Merging & LoRA Composition Exploits
Exploiting model merging techniques (TIES, DARE, linear interpolation) and LoRA composition to introduce backdoors through individually benign model components.
Model Merging Safety Implications
Analysis of how model merging techniques (TIES, DARE, SLERP) affect safety properties and alignment.
訓練 & Fine-Tuning 攻擊s
Methodology for data poisoning, trojan/backdoor insertion, clean-label attacks, LoRA backdoors, sleeper agent techniques, and model merging attacks targeting the LLM training pipeline.
模型 Merging Risks
安全 risks in model and adapter merging workflows -- how merging adapters from untrusted sources can introduce vulnerabilities, exploit merge algorithm properties, and cause safety property loss through TIES, DARE, SLERP, and linear interpolation.
模型 Merging 安全 Analysis
安全 implications of model merging techniques (TIES, DARE, SLERP) including backdoor propagation and safety property degradation.
模型 Merging 攻擊 Surface Analysis
安全 analysis of model merging techniques including TIES, DARE, and SLERP for injecting malicious capabilities.
模型 Merging 安全 Implications
安全 analysis of model merging techniques and potential for backdoor propagation through merged models.
實驗室: 模型 Merging 安全 Analysis
Analyze security implications of model merging techniques and test for backdoor propagation through merged model weights.
模型 Merging Backdoor Propagation
Demonstrate how backdoors propagate through model merging techniques like TIES, DARE, and spherical interpolation.
進階訓練漏洞
AI 訓練中的進階安全威脅——涵蓋聯邦學習攻擊、模型合併風險、水印移除、合成資料投毒、遺忘攻擊與持續學習漏洞。
模型合併與 LoRA 組合攻擊
利用模型合併技術（TIES、DARE、線性內插）與 LoRA 組合，透過個別無害的模型元件引入後門。
模型 Merging Safety Implications
Analysis of how model merging techniques (TIES, DARE, SLERP) affect safety properties and alignment.