CROSS-MAP 框架：推理前替换语义，降低 LLM 隐私泄露风险
LLMs as Linguistic Chameleons: Decoupling Semantics and Structure for Privacy-Preserving Communication
担心发去 API 的内容被还原的话看看这篇，把输入语义整个换掉再让 LLM 推理，重构不回来还不掉点。
随着 LLM API 进入隐私敏感场景，现有方法为保任务效果而保留原始语义，导致文本可被攻击重构。论文提出 CROSS-MAP 框架，在推理前把私密输入映射到另一个语义域，推理后再恢复对应输出。本地模型用多目标优化训练，映射阶段最大化语义差异，恢复阶段最小化语义不一致。实验显示 CROSS-MAP 在多种攻击设置下降低重构成功率，同时任务效果优于现有基线。
LLMs as Linguistic Chameleons: Decoupling Semantics and Structure for Privacy-Preserving Communication
As Large Language Model (LLM) APIs become increasingly integrated into privacy-sensitive workflows, ensuring inference-time privacy without compromising task utility remains a major challenge. Existing approaches preserve most of the original semantic content to maintain downstream performance, but this also leaves exploitable cues for reconstructing the original text. This work investigates semantic decoupling, which replaces original semantics with alternative content while preserving the structure needed for LLM reasoning. Based on this idea, we propose CROSS-MAP, a bidirectional framework that maps private inputs into a different semantic domain before inference and recovers the corresponding outputs afterward. Local models are trained with multi-objective optimization to maximize semantic divergence in the mapping stage while minimizing semantic inconsistency in the recovery stage. Experiments show that CROSS-MAP reduces reconstruction success across multiple attack settings while outperforming existing baselines in utility.