#对抗测试

3 stories · 0 in 7 days · 0 in 30 days

AI安全 1 DeepSeek-R 1 GPT-5-Mini 1 MiniMax-M2.5 1 Penumbra 1 动态规划 1 合规 1 测试策略 1

Stories, products and papers tagged "对抗测试", newest first.

Oct 6

Papers Builders09:26

Chinese contentPenumbra：针对监管义务的样本高效对抗搜索方法

这篇论文做了个叫 Penumbra 的合规测试工具，能用极少的生成次数找出刚踩到监管红线的回复对，金融和医疗场景挺实用。

#Penumbra #对抗测试 #合规 #监管义务

arXiv cs.AI@Anthony RhodesSource

Sep 15

Papers BuildersSelected13:10

Adversarial Testing of Automated Program Repair Agents

For developers, this matters because it shows that current LLM-based automated program repair tools can be tricked into fixing bugs in a way that introduces new security vulnerabilities.

#adversarial testing #automated program repair #security vulnerabilities #LLM

arXiv: DeepSeek@Fares Trad et al. (5)Source

Aug 31

Papers09:31

Chinese content最优对抗测试：从 dishonest 测试者获取诚实结果

这篇论文教你如何设计测试策略，即使有人用 AI 作弊也能得到真实结果。

#AI安全 #动态规划 #对抗测试 #测试策略

arXiv cs.AI@Owen Cox et al. (3)Source