Get Started
Home
Topics
Search
Library
Research questionHow can AI-generated text detectors expose evidence users can inspect and act on?Binary labels and confidence scores do not show which properties of a passage drove a detector’s decision. Users therefore need evidence that connects the prediction to observable textual signals and can be examined for faithfulness.
AI
Evaluation & Benchmarks
Machine Learning
Mechanistic Interpretability
Natural Language Processing
Latest papersRecent research connected to this question, newest first.NOTAI.AI: Explainable Detection of Machine-Generated Text via Curvature and Feature AttributionThe described system combines conditional-probability curvature, a neural detector score, stylometric features, and readability features in an XGBoost meta-classifier. It uses TreeSHAP contributions and natural-language explanations, with evidence from a category-balanced RAID subset containing human-written, clean AI-generated, and attacked AI-generated text; reported results are limited to a held-out within-subset test split and automatic explanation-faithfulness judgments.research paper · Sep 3, 2026
Related questions
How can AI-generated text detectors remain reliable under distribution shift?How can image forensics detect manipulated text, locate altered regions, and ground explanations in evidence?How can we distinguish AI assistance from AI authorship when the prompting dialogue is unavailable?How do we evaluate whether scientific agents make justified discoveries from data rather than reproduce known analyses?