Skip to content

Dataset · stepfun-ai

PaCoRe data for parallel reasoning

stepfun-ai/PaCoRe-Train-8k

About 8K text examples for training models in parallel coordinated reasoning within PaCoRe.

What's inside

The dataset contains text examples for training PaCoRe, an approach that launches parallel reasoning branches and coordinates them. It is designed to scale test-time compute through width rather than just depth. Data is in English, with a size ranging from 1K to 10K examples.

What you can do with it

  • Model SFT on PaCoRe traces
  • Training parallel reasoning
  • Experiments with test-time compute
Tasks
Text generation
Languages
English
Size
1K–10K rows
Downloads this month
881
Likes
81

How to load it

from datasets import load_dataset

ds = load_dataset("stepfun-ai/PaCoRe-Train-8k")