Skip to content

Dataset · bespokelabs

Questions, reasoning, and answers for reasoning fine-tuning

bespokelabs/Bespoke-Stratos-17k

17k examples of questions, reasoning chains, and answers from DeepSeek-R1 for fine-tuning reasoning models.

What's inside

The dataset contains questions, reasoning traces, and answers. It was compiled via fine-tuning distillation of DeepSeek-R1 data and replicates the improved Berkeley Sky-T1 pipeline. Bespoke-Stratos-32B (fine-tune Qwen-2.5-32B-Instruct) and Bespoke-Stratos-7B (fine-tune Qwen-2.5-7B-Instruct) were trained on this data. The language is English.

What you can do with it

  • fine-tune a 7B/32B reasoning model
  • fine-tune Qwen-2.5 LoRA for reasoning chains
  • compare with the Sky-T1 pipeline
Languages
English
Size
10K–100K rows
Downloads this month
38.8K
Likes
366

Models trained on it

How to load it

from datasets import load_dataset

ds = load_dataset("bespokelabs/Bespoke-Stratos-17k")