Skip to content

Dataset · miromind-ai

MiroVerse v0.1 deep research data

miromind-ai/MiroVerse-v0.1

English SFT and DPO data for training and reproducing the MiroThinker-v0.1 deep research agent on Qwen3.

What's inside

MiroVerse v0.1 is a growing deep research dataset compiled in a reproducible full-trajectory format. Version one includes SFT and DPO data compatible with the MiroTrain framework. Language is English, size is 100k to 1M examples.

What you can do with it

  • Train a deep research agent via SFT
  • Tune DPO for MiroThinker-v0.1
  • Reproduce the benchmark on Qwen3
Tasks
Question answering
Languages
English
Size
100K–1M rows
Downloads this month
543
Likes
242
Access
on request on Hugging Face

Models trained on it

How to load it

from datasets import load_dataset

ds = load_dataset("miromind-ai/MiroVerse-v0.1")