Overview
ChenlongDeng/ADAPT-Qwen2-7B-CAIL2018-step-8765 is a 7.6 billion parameter model built upon the Qwen2-7B architecture, specifically fine-tuned for legal judgment prediction. This model is the result of research detailed in the paper "Enabling Discriminative Reasoning in LLMs for Legal Judgment Prediction" by Chenlong Deng, Kelong Mao, Yuyao Zhang, and Zhicheng Dou. It leverages the CAIL2018 dataset to enhance its ability to perform legal reasoning tasks.
Key Capabilities
- Legal Judgment Prediction: Analyzes case descriptions to predict potential criminal charges, applicable legal articles, and sentencing.
- Discriminative Reasoning: Designed to enable more nuanced and discriminative reasoning within legal contexts.
- Structured Prompting: Supports four distinct prompt formats (ADAPT Reasoning, Ask, Article, Sentencing factors) to guide the model in specific legal analysis tasks.
- Training Trajectories: The project also releases 80,141 training trajectories from the CAIL2018 dataset, which can be valuable for further research.
Good For
- Legal AI Applications: Ideal for developers building systems that require automated legal analysis and judgment prediction.
- Academic Research: Useful for researchers studying the application of large language models in the legal domain, particularly for discriminative reasoning.
- Case Analysis: Can assist in preliminary analysis of legal cases by identifying relevant laws and potential outcomes based on provided case details.