Julian2002/PDP-Qwen3-8B-SFT
Julian2002/PDP-Qwen3-8B-SFT is an 8 billion parameter Qwen3-based language model fine-tuned for prosecution decision prediction (PDP). This model specializes in generating structured prosecutorial reasoning and final decisions based on suspect, procedural, and factual information. It is specifically trained to classify decisions into four distinct legal outcomes, making it optimized for legal reasoning tasks within the Chinese judicial context.
Loading preview...
Model Overview
Julian2002/PDP-Qwen3-8B-SFT is an 8 billion parameter language model built upon the Qwen/Qwen3-8B architecture. It has undergone supervised fine-tuning (SFT) specifically for prosecution decision prediction (PDP) experiments, utilizing the PDP-Bench / pdp2k_rq3_sft dataset.
Key Capabilities
- Prosecution Decision Prediction: The model is trained to analyze suspect, procedural, and factual information to predict prosecutorial decisions.
- Structured Reasoning Generation: It generates detailed prosecutorial reasoning in a predefined format, including sections for applicable laws, review analysis, and final conclusions.
- Specific Decision Space: The model's output is constrained to four distinct Chinese legal labels:
起诉(prosecution),相对不起诉(relative non-prosecution),法定不起诉(statutory non-prosecution), and存疑不起诉(non-prosecution due to insufficient evidence).
Use Cases
This model is particularly well-suited for applications requiring automated legal reasoning and decision support within the Chinese judicial system, specifically for tasks related to prosecution decision-making. Its structured output format makes it valuable for integrating into legal tech platforms that require explainable AI for judicial processes.