Stratified-LLM-Subsets-100K-1M-Scale
PublicStratified LLM Subsets delivers diverse training data at 100K-1M scales across pre-training (FineWeb-Edu, Proof-Pile-2), instruction-following (Tulu-3, Orca AgentInstruct), and reasoning distillation (Llama-Nemotron). Embedding-based k-means clustering ensures maximum diversity across 5 high-quality open datasets.
datadatasetdeep-learningdiversityfine-tuningfinetuningfinetuning-llmsinstruction-followinginstruction-tuningllm
Creat：2025-09-30T16:35:20
Update：2025-10-06T03:43:33
https://amanpriyanshu.github.io/Stratified-LLM-Subsets-100K-1M-Scale/
1
Stars
0
Stars Increase