EASE: Entropy-Adaptive Distribution Shaping for Evading AI-generated Text Detectors

Key points

• We observe that perturbing next-token logits or adjusting sampling temperature can reduce detection performance, providing a clear signal of detector vulnerability to decoding-time distribution changes.1
• Building on this observation, we propose EASE (Entropy-Adaptive Distribution Shaping for Evasion), a training-free and detector-agnostic framework for evading AIGT detectors.1
• EASE computes predictive entropy directly from the source LLM's next-token distribution and uses it to adapt both logit perturbation and sampling temperature, without detector feedback or model fine-tuning.1
• Experiments across three source LLMs and multiple detectors demonstrate consistent reductions in detection performance, with negligible degradation in text quality and negligible inference overhead.1

Sources (1)

• [1]EASE: Entropy-Adaptive Distribution Shaping for Evading AI-generated Text Detectors
  arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 7, 12:41 PM

  AI-generated text (AIGT) detection can be sensitive to the decoding choices of the source large language model (LLM).

  We observe that perturbing next-token logits or adjusting sampling temperature can reduce detection performance, providing a clear signal of detector vulnerability to decoding-time distribution changes.

Extractive summary: sentences quoted from the sources.