Ensemble techniques in validating chain of reasoning
A machine-learning pipeline that checks whether a language model's step-by-step reasoning holds up, built with PyTorch and GPT-3.5 Turbo through systematic prompt-engineering experiments.
+26% F1 score and a 30% accuracy gain over baseline.
- Python
- PyTorch
- TensorFlow
- NumPy
- Pandas
- Matplotlib
- OpenAI API