PaperScholar: An AI Academic Writing Mentor System Designed for Researchers

PaperScholar is an AI academic writing toolchain for Chinese and English-speaking researchers. Its core positioning is to simulate a strict mentor role, helping researchers produce high-quality academic papers through features like mentor mode, AI tone removal, innovation detection, and rebuttal generation. Its core philosophy is 'Every claim needs evidence; build the structure before writing sentences', aiming to enhance the rigor and standardization of academic writing.

For researchers, especially non-native English speakers, academic writing faces multiple challenges in conception, structure, language, etc. Existing AI writing tools often have 'AI tone' issues such as empty clichés, excessive conjunctions, and lack of empirical support. PaperScholar emerged to address these pain points, simulating a strict mentor role to assist academic writing.

PaperScholar's core philosophy: 'Every claim needs evidence; build the structure before writing sentences'. It emphasizes evidence first (establishing an evidence library with source annotations), structure priority (planning chapter structure and motivation), and mentor perspective (simulating top reviewer standards). The workflow is pipeline-based: Input materials → Ingestion and verification → Evidence library construction → Blueprint design → Writing and rewriting → Post-processing (mentor review, AI tone detection, etc.) → Output files.

PaperScholar包含八大核心功能：

1. Mentor Mode: Simulates reviewers/mentors from top conferences like NeurIPS and CVPR, providing scores and revision suggestions;
2. AI Tone Removal: Detects and eliminates AI-generated features such as empty opening remarks and excessive conjunctions;
3. Innovation Detection: Distinguishes between real innovation and pseudo-innovation (e.g., method migration, parameter tuning, etc.);
4. Rebuttal Generation: Analyzes review comments item by item and generates structured bilingual responses;
5. Evidence Library: Forces claims to link to sources, avoiding AI hallucinations;
6. Blueprint Priority: Plans structure and motivation first to ensure clear logic;
7. Paper Deconstruction: Reverse-learn the structure of top papers;
8. Multi-model Support: Compatible with over 8 models like DeepSeek and OpenAI.

Application scenarios cover journal papers, conference papers, dissertations, project applications, academic competitions, technical reports, etc. The usage methods are flexible: a code-free Gradio web interface, or command-line batch processing—just need an API key and input materials to start the process.

PaperScholar is developed in Python, with a UI built on Gradio, supporting OpenAI-compatible APIs. The project is open-sourced under the MIT license, encouraging the community to contribute new mentor styles, Chinese scenario templates, prompt optimization, etc., to lower the usage threshold and promote continuous improvement.

PaperScholar represents a new paradigm for AI-assisted academic writing—it does not replace human thinking, but enhances researchers' writing abilities by simulating a strict mentor. It helps researchers establish rigorous thinking, produce high-quality papers, and respond to review scrutiny. In the context of AI popularization, PaperScholar provides an example for the responsible use of AI in assisting academic writing.