RESEARCH DEMO

SRL

StealthRL

AI-text detectors are increasingly used in high-stakes settings, yet their robustness to meaning-preserving adversarial rewriting remains uncertain. StealthRL is an RL framework for stress-testing AI-text detectors with adaptive paraphrase attacks. StealthRL trains a paraphrase policy against a detector ensemble while preserving semantic content, then evaluates transfer to held-out detector families.

Research preview. Do not paste private text. Outputs are for robustness exploration, not detector-evasion guarantees.

Text to paraphrase

Loading quota...
Generation config

Defaults match the paper evaluation setting.

Use API key for extended access

Stored only in this browser. Public requests use the daily anonymous quota.

Ready.