Unverified50% confidenceFactExact time
Early GCG (Greedy Coordinate Gradient) adversarial attacks required thousands of steps of gradient optimization.
1
Sources
50%
Confidence
Long-term
Relevance
9/15/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedIn 2023, a Carnegie Mellon University research team discovered a gradient-based automated jailbreak method (the GCG attack), which can bypass the safety protections of almost all mainstream models by appending seemingly meaningless suffix strings to prompts.68% similarUnverifiedThe core methodology adopted in this optimization is 'follow the bottleneck,' which systematically identifies and eliminates performance bottlenecks one by one.62% similarUnverified撞库攻击利用已知有效凭证批量尝试登录，成功率远高于随机猜测，约0.1%-2%的撞库尝试能够成功60% similarUnverifiedRetrieval-Augmented Generation (RAG), long-context modeling optimization, and targeted fine-tuning for academic scenarios may become key directions for breaking through the current performance bottleneck in evidence extraction.60% similarUnverified潜在空间寻优是高维黑箱优化问题，常用策略包括贝叶斯优化、进化算法和基于梯度的方法59% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/922739API
curl https://kongchang.com/api/v1/knowledge/claims/922739MCP
get_claim(id=922739)