Search papers, labs, and topics across Lattice.
This study explores evasion attacks against state-of-the-art AI writing detectors, revealing that structural shifts in generated text can effectively bypass adversarial fine-tuning. By demonstrating that pushing text out of the detector's training distribution is significantly more effective than attempting to mimic it, the authors introduce two novel attack families that achieve up to 50x higher fool rates while maintaining naturalness. The findings highlight persistent vulnerabilities in detector families, even those that have undergone adversarial fine-tuning, underscoring the need for improved detection strategies.
Pushing AI-generated text beyond the training distribution can defeat even the most advanced adversarial detectors, revealing a critical vulnerability in current AI safety measures.
We investigate which language model evasion attacks survive state-of-the-art adversarial fine-tuning, developing strategies that sweep the top 5 positions on the ELOQUENT 2026 Voight-Kampff leaderboard. While adversarial fine-tuning trivially closes the 2025 winning evasion recipes, we uncover a fundamental asymmetry in detector vulnerability: pushing generated text out of the detector's training distribution reliably defeats adversarial detection, whereas pulling it into the distribution (e.g., mimicking human training data) fails completely. Exploiting this, we introduce two novel out-of-distribution attack families - cross-decade register attacks and modernist stream-of-consciousness form. Both strategies easily bypass adversarial closure, achieving up to approximately 50x higher fool rates than previous methods while preserving naturalness. Furthermore, experiments show that the obvious deployer countermeasure (augmenting training data with period prose) fails to close the vulnerability. Our findings show that the tested detector families, including adversarially fine-tuned ones, exhibit persistent vulnerabilities under structural out-of-distribution shifts, a mechanism that directly powers our leading competition performance.
