Research map / Adversarial robustness and security / Attacks on text and graphs
Attacks on Machine-Text Detectors Retain Stylistic Fingerprints
Rafael Alberto Rivera Soto, Barry Y. Chen, Nicholas Andrews
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| optimize the generator with DPO against FastDetectGPT | Mistral/Qwen/Nemo generators · Reddit | AUROC (FastDetectGPT): 65 -> 35 | Hurt |
| optimize the generator with DPO against FastDetectGPT | generators, tested with StyleDetect · Reddit | AUROC (StyleDetect): 97 -> 96 | No clear effect |
| DPO targeting StyleDetect itself | generators · Reddit | AUROC (StyleDetect): 97 -> 96 | No clear effect |
| apply TinyStyler rewriting to machine text | machine text · Reddit | AUROC (Binoculars): 79 -> 92 | Hurt |
| apply TinyStyler rewriting to machine text | machine text · Reddit | AUROC (StyleDetect): 98 -> 87 | Mixed |
| apply DIPPER paraphrasing | machine text · Reddit | AUROC (StyleDetect): 98 -> 89 | Mixed |
| apply GPT-4o-mini paraphrasing | machine text · Reddit | AUROC (FastDetectGPT): 71 -> 59 | Improved |
| replace TinyStyler with the proposed style-aware paraphraser | style transfer · Reddit 180 author pairs | CISR stylistic similarity: 0.71 -> 0.83 | Improved |
| replace TinyStyler with the proposed style-aware paraphraser | style transfer · Reddit 180 author pairs | SBERT semantic similarity: 0.74 -> 0.83 | Improved |
| add DPO post-training to the style-aware paraphraser (ablation) | style-aware paraphraser · Reddit | AUROC(1) best detector vs sample size | Improved |
| increase number of target exemplars M or paraphrases P (ablation) | style-aware paraphraser · Reddit | stylistic similarity to target | Improved |
| increase documents per author from N=1 to N=50 | Ours on newer generators · Reddit | AUROC(1) strongest detector: 0.5452 -> 0.8816 | Hurt |
Closest papers by meaning
- Few-Shot Detection of Machine-Generated Text using Style Representations (ICLR 2024 · similarity 0.88)
- Imitate Before Detect: Aligning Machine Stylistic Preference for Machine-Revised Text Detection (AAAI 2025 · similarity 0.87)
- Adversarial Paraphrasing: A Universal Attack for Humanizing AI-Generated Text (NeurIPS 2025 · similarity 0.86)
- Advancing Machine-Generated Text Detection from an Easy to Hard Supervision Perspective (NeurIPS 2025 · similarity 0.85)
- DALD: Improving Logits-based Detector without Logits from Black-box LLMs (NeurIPS 2024 · similarity 0.85)
- Log-Likelihood, Simpson’s Paradox, and the Detection of Machine-Generated Text (NeurIPS 2026 · similarity 0.84)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
