A synthetic reasoning dataset spanning Mathematics, Coding, and Healthcare — designed for training and evaluating LLMs on step-by-step reasoning across domains.
1,500 samples — 500 per domain (math, coding, healthcare)
Step-by-step reasoning — every answer follows a numbered Step 1: → Final Answer: format
Cross-domain overlap — ~15% of each domain intentionally overlaps with each other domain (e.g., biostatistics… See the full description on the dataset page:
https://huggingface.co/datasets/crevious/Tri-NL.
