This supervised fine-tuning (SFT) dataset is built from the 2024 Florida International University (FIU) Physics Graduate Qualifying Exam, pairing questions in classical mechanics, electromagnetism, and statistical physics with full step-by-step solutions.
Potential Use Cases
- Supervised Fine-Tuning for Scientific Reasoning: Trains LLMs to carry out multi-step, physically grounded derivations — selecting the right framework, applying boundary conditions, and checking limits and units — instead of pattern-matching to a final answer.
- Expert-Level Evaluation Benchmarking: Provides doctoral-qualifying difficulty for measuring model capability by subfield (classical mechanics, electromagnetism, statistical physics), allowing final-answer accuracy to be scored separately from process correctness.