跳到正文
The Decoder· Manuel Uth·· 5 天前AI 评分62

Anthropic 称其防御措施将 Claude 推理 token 窃取量降低高达 99%

Anthropic says it cut Claude reasoning token theft by up to 99% in fight against industrial-scale distillation

AI 导读

Anthropic 宣布其新防御措施已将 Claude 模型内部推理链(chain-of-thought)的 token 窃取量减少 90% 至 99%,以应对大规模蒸馏攻击。

来源:The Decoder · the-decoder.com