ConceptNemotron Post Training Dataset v2
Nemotron-Post-Training-Dataset-v2
An open-source dataset released by NVIDIA for post-training. It contains multiple splits covering mathematics, STEM, code, and general conversation. Some samples include reasoning chains and code blocks. It is built under the CC BY license.
Timeline (last 90 days)
Oct 13
In the English split of Nemotron-Post-Training-Dataset-v2, the largest share is the mathematics and STEM portion; these samples are long complete reasoning processes, and a single item is far longer than a customer service ticket.
Unverified50%