UTS bypasses ELOQUENT 2026 Voight-Kampff detectors with structural shifts

Researchers from UTS developed evasion strategies that sweep the top 5 positions on the ELOQUENT 2026 Voight-Kampff leaderboard by exploiting a fundamental asymmetry in detector vulnerability. The study finds that pushing generated text out of the detector's training distribution reliably defeats adversarial detection, whereas mimicking human data fails.

• Two novel out-of-distribution attack families were introduced: cross-decade register attacks and modernist stream-of-consciousness form.
• These strategies achieve up to approximately 50x higher fool rates than previous methods while preserving naturalness.
• Experiments show that augmenting training data with period prose fails to close the vulnerability.

The findings demonstrate that tested detector families, including adversarially fine-tuned ones, exhibit persistent vulnerabilities under structural out-of-distribution shifts.