High-Fidelity Chain-of-Thought Distillation • Verified Teacher Forcing • Rich LaTeX
This dataset provides high-quality, step-by-step Chain-of-Thought (CoT) reasoning generated by the OpenAI gpt-oss-120bmodel (via OpenRouter/Novita). It is specifically curated for Math Reasoning SFT on small models (0.5B–7B), using a rigorous Verified Teacher Forcing pipeline to eliminate hallucinations.
The dataset is:
✔ Verified: Only samples where the… See the full description on the dataset page:
https://huggingface.co/datasets/HAD653/OpenMath-Easy-CoT-Verified-120B.
