mayiwen/PaperAudit_Qwen3_14B_sft_rl
The mayiwen/PaperAudit_Qwen3_14B_sft_rl model is a 14 billion parameter language model based on Qwen3, specifically optimized for academic paper error detection and automated review tasks. It leverages Supervised Fine-Tuning (SFT) and Reinforcement Learning from Human Feedback (RLHF) to achieve high accuracy in analyzing complex academic content. With a 40,960-token context length, it excels at processing complete academic papers, understanding complex concepts, and providing professional-grade feedback.
Loading preview...
Model Overview
PaperAudit_Qwen3_14B_sft_rl is a 14 billion parameter model built upon the Qwen3 architecture, specifically engineered for academic paper error detection and automated review. It has undergone Supervised Fine-Tuning (SFT) and Reinforcement Learning from Human Feedback (RLHF) to align with human preferences and enhance review quality.
Key Capabilities
- Specialized Optimization: Uniquely optimized for identifying errors and providing professional reviews in academic papers.
- Long Context Support: Processes entire academic papers with a 40,960-token context length.
- Deep Understanding: Capable of comprehending complex academic concepts, writing norms, and performing critical analysis.
- High Accuracy: Excels at identifying subtle and complex academic errors, offering detailed and professional review comments.
- Comprehensive Feedback: Provides not only error identification but also constructive improvement suggestions.
Training and Performance
The model was trained on the PaperAudit_Dataset, which includes academic papers, structured content, synthetic error data, and human review feedback. Compared to smaller 3B and 8B models, the 14B version offers significantly higher accuracy, deeper analysis, and stronger reasoning capabilities for academic tasks.
Suitable Scenarios
- High-quality academic journal review.
- Thesis and dissertation quality assessment.
- Academic conference paper review.
- Any scenario demanding extremely high review quality and deep academic analysis.