chrononauts
chrononauts-sft-filteredchrononauts-think-sft-filteredchrononauts-dpo-filteredchrononauts-retrieval-sft-202407-500
Chrononauts Retrieval-Grounded SFT (202407, n=500)
500 Think-SFT examples that teach a language model to ground its answers in retrieved documents rather than relying on parametric knowledge. Schema is byte-identical to allenai/Dolci-Think-SFT-7B so it can be folded into existing Think-SFT training pipelines without modification.
Proof-of-concept dataset for the Chrononauts project, targeting the 202407 model cutoff.
Schema
Three top-level fields per example:
{… See the full description on the dataset page: https://huggingface.co/datasets/mattwang123/chrononauts-retrieval-sft-202407-500.