mgoin/open-perfectblend-glm5.2-regen

open-perfectblend-glm5.2-regen On-policy regeneration of the full mlabonne/open-perfectblend with GLM-5.2-FP8, built to train speculative-decoding drafters (dspark / DFlash). 1,420,229 conversations in ShareGPT-style {id, conversations: [{from, value}], source}. Each assistant turn was regenerated by GLM-5.2-FP8 conditioned on the preceding, already-regenerated turns — deepspec-style per-turn on-policy regeneration, up to 8k tokens per turn. Original human turns are preserved.… See the full description on the dataset page: https://huggingface.co/datasets/mgoin/open-perfectblend-glm5.2-regen.