sft-100k-mixture – Dataset by TeamClaude | AlphaNeural AI

You can deploy this model and start earning money today!

TeamClaude/
sft-100k-mixture

like

0

text-generation

instruction-tuning

Views

No views yet

SFT 100K Mixture Dataset

A curated 100,000-row SFT mixture dataset for instruction-tuning LLMs.

	Mixture Recipe

Component Source Dataset Rows Purpose

Logic Core (60%) PrimeIntellect/SYNTHETIC-2-SFT-verified 60,000 Reasoning, math, coding with verified DeepSeek-R1 traces

Conversational Glue (30%) Magpie-Align/Magpie-Llama-3.1-Pro-300K-Filtered 30,000 Multi-turn fluidity, tone alignment

Human Element (10%) allenai/WildChat-4.8M 10,000 Real-world prompts… See the full description on the dataset page: https://huggingface.co/datasets/TeamClaude/sft-100k-mixture.