Research map / Large language models / Language, safety and interpretability
Beyond Raw Detection Scores: Markov-Informed Calibration for Boosting Machine-Generated Text Detection
Chenwang Wu, Yiu-ming Cheung, Shuhai Zhang, Bo Han, Defu Lian
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 |
|---|---|---|---|
| add Markov-informed calibration (MRF + positional weighting) | Likelihood · Essay (avg) | AUROC: 96.17 -> 97.79 | Improved |
| add Markov-informed calibration | DetectGPT · Essay (avg) | AUROC: 44.09 -> 91.95 | Improved |
| add Markov-informed calibration | Log-Rank · DetectRL (avg) | AUROC: 72.06 -> 83.29 | Improved |
| add Markov-informed calibration | FastGPT · DetectRL (avg) | AUROC: 60.97 -> 61.68 | No clear effect |
| add Markov-informed calibration | FastGPT · Essay (avg) | AUROC: 69.08 -> 79.92 | Mixed |
| add Markov-informed calibration | DNA-GPT · DetectRL (avg) | AUROC: 67.79 -> 69.75 | Improved |
| remove MRF layer or positional weighting (ablation) | Markov-informed calibration · Essay | AUROC: significant drop when either removed | Improved |
| use three-layer neural network for score calibration instead of MRF (ablation) | Likelihood · cross-LLM detection | AUROC: generalization drops significantly | Hurt |
| train detector on mixed text instead of original (ablation) | detectors · mixed-text detection | AUROC: no number in text | No clear effect |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- DetectGPT — DetectGPT: Zero-Shot Machine-Generated Text Detection using Probability Curvature (ICML 2023) · better on 2 of 2 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Log-Likelihood, Simpson’s Paradox, and the Detection of Machine-Generated Text (NeurIPS 2026 · similarity 0.88)
- Advancing Machine-Generated Text Detection from an Easy to Hard Supervision Perspective (NeurIPS 2025 · similarity 0.87)
- Detecting Machine-Generated Texts by Multi-Population Aware Optimization for Maximum Mean Discrepancy (ICLR 2024 · similarity 0.86)
- Training-free LLM-generated Text Detection by Mining Token Probability Sequences (ICLR 2025 · similarity 0.86)
- DALD: Improving Logits-based Detector without Logits from Black-box LLMs (NeurIPS 2024 · similarity 0.85)
- Fast-DetectGPT: Efficient Zero-Shot Detection of Machine-Generated Text via Conditional Probability Curvature (ICLR 2024 · similarity 0.85)
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.
