A mix of publicly-available instruction following/conversational datasets
Token Counts
Source Datasets
ai2-adapt-dev/flan_v2_converted - FLAN v2 converted to a modern chat format
allenai/tulu-3-sft-personas-instruction-following - Instruction following dataset made for Tulu 3
simplescaling/s1K-1.1 - S1K with Deepseek R1 thinking and responses
simplescaling/s1K-claude-3-7-sonnet - S1K with Claude 3.7 thinking and responses… See the full description on the dataset page: https://huggingface.co/datasets/allura-org/inkstructmix-v0.2.
