Confirmed
Hugging Face releases FineDNA, a dataset for teaching LLMs DNA
FineDNA includes over 300B tokens of genomes and DNA-rich text for midtraining, about 200K SFT conversations and about 98K verifiable RL prompts. The dataset and code are open, and the data was used to build the Carbon-X model on Qwen3-14B.
Every outlet that ran it 1 article
r/LocalLLaMACommunityLead
FineDNA: the largest dataset to teach any LLM DNA (midtraining, SFT, RL)