1
The Injection Paradox: Brand-Level Suppression in Safety-Trained LLM Recommendations via RAG Context Injection
一篇揭示RAG-LM安全训练中“注入悖论”的论文:注入逆向抑制品牌推荐,带来新的安全思考。
arXiv:2606.09204v1 Announce Type: new Abstract: We present a reproducible failure mode of safety training in RAG-based LLM recommendation -- the Injec…