From Leaky Thoughts to Private Reasoning: Controlling What LRMs Say to Themselves 文章

ArXiv CS.CL2026-06-01NEWSen作者: Haritz Puerto, Haonan Li, Xudong Han, Timothy Baldwin, Iryna Gurevych
查看原文 →
关系图谱

From Leaky Thoughts to Private Reasoning: Controlling What LRMs Say to Themselves · 相关技术

相关技术

LoRAStaged Decodingprompt injection attacksreasoning tracesLarge Reasoning Models