AI Paper Atlas

Research map / Large language models / Language, safety and interpretability

Multi-level Style Preference Optimization: An Adaptive Detection Framework for Human-Machine Hybrid Text

Zehao Wang, Lianwei Wu, Wenbo An, Hang Zhang, Yaxiong Wang

AAAI 2026 · Large language models › Language, safety and interpretability · approach: Zero-shot machine-text detection · 0 citations (OpenAlex)

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 changedWhereMeasuredEffect
add phrase-level and lexical-level losses to sequence-level loss (ablation)MSPO · GPT-3.5 polished CSAUROC: 0.9175 -> 0.9774Improved
add lexical-level KL constraint (ablation)MSPO · GPT-3.5 polished CSAUROC: 0.9227 -> 0.9774Improved
add phrase-level n-gram hinge loss (ablation)MSPO · GPT-3.5 polished CSAUROC: 0.9350 -> 0.9774Improved
add complexity-based dynamic beta (ablation)MSPO · GPT-3.5 polished CSAUROC: 0.9429 -> 0.9774Improved
use MSPO instead of ImBD on rewritten textMSPO · GPT-3.5 rewritten, PHXAUROC: 0.7099 -> 0.8214Improved
use MSPO instead of ImBD on polished textMSPO · XSum polished (4 LLM avg)AUROC: 0.8208 -> 0.8809Improved
use MSPO instead of ImBD on LLaMA3.1 rewritten newsMSPO · XSum rewritten, LLaMA3.1AUROC: 0.5932 -> 0.7715Improved
use MSPO instead of ImBD on purely generated textMSPO · GPT-3.5 generated, CSAUROC: 0.9997 -> 0.9907No clear effect
use MSPO instead of ImBD on creative-writing rewritesMSPO · WritingPrompts rewritten, DeepSeek/GPT-4oAUROC: GPT-4o 0.6216 -> 0.7094; GPT-3.5 0.6206 -> 0.7625; LLaMA3.1 0.7082 -> 0.7329Mixed

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy