chauben/advisorai-qwen2.5-14b-stevens
chauben/advisorai-qwen2.5-14b-stevens is a 14.8 billion parameter language model fine-tuned from Qwen/Qwen2.5-14B-Instruct by Nitin Chaube. This model is specifically optimized as AdvisorAI, an academic advising assistant for Stevens Institute of Technology. It excels at answering student-style questions about Stevens, covering courses, programs, faculty, admissions, and general advising. The model leverages QDoRA and NEFTune for efficient fine-tuning and provides helpful, markdown-formatted responses.
Loading preview...
AdvisorAI: Stevens Institute Academic Advisor
chauben/advisorai-qwen2.5-14b-stevens is a specialized large language model, fine-tuned from the robust Qwen/Qwen2.5-14B-Instruct base model. Developed by Nitin Chaube, this 14.8 billion parameter model is designed to function as AdvisorAI, an academic advising assistant tailored for the Stevens Institute of Technology.
Key Capabilities
- Academic Advising: Provides detailed answers to student inquiries about Stevens Institute of Technology.
- Domain-Specific Knowledge: Covers a wide range of topics including:
- Courses and their prerequisites
- Academic programs and degree requirements
- Faculty information (based on training data)
- Admissions processes, financial aid, and campus life
- Formatted Responses: Generates helpful, markdown-formatted advice, often citing specific course codes and requirements.
Training and Specialization
This model was fine-tuned using QDoRA (4-bit NF4 + DoRA r=64 + rsLoRA) and NEFTune (α=5) techniques, enhancing its performance on the specific academic advising domain. The training dataset comprised nearly 80,000 examples derived from Stevens-related sources and LLM-assisted Q&A generation, primarily focusing on course and general advising queries. The final checkpoint is a merged model with the DoRA adapter fused into the base weights, ensuring direct usability.