AIGC-Shed Paper AIGC Reduction v3

Problem Framing

aigc-shed targets the AIGC suspicion issue in revised papers. It does not rely on simple synonym swaps; instead, it adjusts sentence structure and word-frequency patterns against detector preferences. The default goal is to bring the AIGC rate below 15%, while extreme mode targets around 0.5%, avoids three consecutive sentences of similar length, splits long sentences into two or three parts, and inserts ultra-short 3-8 character segments.

How It Works

The skill loads detector-specific strategies such as CNKI, VIP, Wanfang, GPTZero, ZeroGPT, or generic modes, and selects light, moderate, deep, or extreme intensity. Core actions include:
- Function-word reduction: lowers template-like phrasing by tightening common particles.
- Rhythm disruption: splits long sentences and inserts short clauses to break predictable cadence.
- Term protection: preserves user-specified terms so technical meaning does not drift.
- Multi-model relay: enables a pipeline-based rewrite, recommended for targets below 10%.
It also tracks feedback: after every five detection records, it adjusts technique weights based on successful and failed cases; if the user does not provide the post-rewrite AIGC rate, it asks for that data.

Boundaries

This skill is best for Chinese papers, reports, or technical prose where AIGC suspicion must be reduced. It is not a substitute for factual accuracy, data integrity, or original reasoning. The term list needs careful maintenance, and extreme mode can make text sound more human, but it does not fix weak research logic.