研究人员推出了一种名为Fair-ASR的新协议，用于评估GPT-5等大型语言模型的越狱攻击。该方法通过使用共享的目标调用预算来标准化比较，解决了先前依赖FLOPs或忽略预算限制的评估方法的局限性。研究发现，传统的攻击方法仍然具有竞争力，并引入了一种新颖、预算高效的ReCode攻击，该攻击在GPT-5上取得了85%的成功率，且攻击者调用次数显著减少。 AI
影响 为LLM安全评估建立了更严格的标准，可能导致更强大的越狱防御措施。
排序理由 该集群报道了一篇介绍LLM新评估协议和攻击方法的新学术论文。
在 Hugging Face Daily Papers 阅读 →
AI 生成摘要 · Google Gemini · 来自 3 个来源。 我们如何撰写摘要 →
完整方法见我们的编辑标准。
arXiv:2511.04707v2 Announce Type: replace-cross Abstract: Recent advances in long-context language models (LMs) have enabled million-token inputs, expanding their capabilities across complex tasks like computer-use agents. Yet, the safety implications of these extended contexts r…
arXiv:2608.17360v1 Announce Type: cross Abstract: Reliable jailbreak evaluation is essential for assessing LLM safety, but most existing studies rely solely on attack success rate (ASR) without accounting for its dependence on attack budgets, resulting in unfair comparisons acros…
Reliable jailbreak evaluation is essential for assessing LLM safety, but most existing studies rely solely on attack success rate (ASR) without accounting for its dependence on attack budgets, resulting in unfair comparisons across methods. Existing compute-aware evaluations redu…