elvis@omarsar0
42AI 编辑部评分,满分 100
2026-08-11 22:42· 14分钟前
AI 导读

推荐阅读。 “这表明,蒸馏推理痕迹可能早已可行,且从未破坏加密。”

Recommended reading.

"This suggests that distilling reasoning traces may have been possible for a long time without ever breaking the cryptography."

Alexander PanfilovWe can finally talk about it: We found a way to extract hidden reasoning of frontier models using a vulnerability in the APIs of every frontier AI company. We v...

来源:elvis · x.com

elvis · @omarsar0 · X·2026-08-11 22:42·14分钟前
AI 导读

推荐阅读。 “这表明,蒸馏推理痕迹可能早已可行,且从未破坏加密。”

Recommended reading.

"This suggests that distilling reasoning traces may have been possible for a long time without ever breaking the cryptography."

Alexander PanfilovWe can finally talk about it: We found a way to extract hidden reasoning of frontier models using a vulnerability in the APIs of every frontier AI company. We v...

来源:elvis· x.com