Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
CISPA Helmholtz Center for Information Security, R-l2italic_R - italic_l 2 metrics. Figure 6: Performance of AQA models on FLEX dataset. Rβl2(Γ100)R-l2(\times 100)italic_R - italic_l 2 ( Γ 100 ) AQA Model Οββπabsent\rho\uparrowitalic_Ο β Rβlβ’2ββπ π2absentR-l2\downarrowitalic_R - italic_l 2 β Single viewyu2021group 0.8069
1
0
3
6
Even the most advanced LLMs like GPT-5.2 and Gemini-3-Pro often fail to recognize and refuse to process harmful content embedded within seemingly harmless tasks.