Search papers, labs, and topics across Lattice.
1
0
3
0
Current AI-text detectors are surprisingly brittle: a novel RL-based paraphrasing attack can reduce their detection ability to near-zero while preserving meaning, even against detectors they weren't trained on.
