Dataset · argilla
7k-example Capybara DPO dataset
argilla/distilabel-capybara-dpo-7k-binarized
Binarized preferences for DPO based on Capybara: multi-turn dialogues for training chat models.
What's inside
A preference dataset for DPO collected via distilabel based on LDJnr/Capybara. This is a 7k-example preview version; v2 promises the full base dataset and responses from stronger models. Language is en, table and text format, size 1K-10K.
What you can do with it
- Train a chat model via DPO
- Conduct RLHF preferences
- Evaluate multi-turn responses
- Tasks
- Question answering, Text generation
- Languages
- English
- Size
- 1K–10K rows
- Downloads this month
- 7.5K
- Likes
- 193
Models trained on it
- kaist-ai/mistral-orpo-capybara-7kText generation 8.9K
- MaziyarPanahi/zephyr-orpo-141b-A35b-v0.1-GGUFText generation 1.4K
- bartowski/zephyr-orpo-141b-A35b-v0.1-GGUFText generation 1.3K
- mradermacher/mistral-7b-orpo-capybara-reproduction-GGUF 1.1K
- mradermacher/Quyen-Plus-v0.1-i1-GGUF 851
- mradermacher/Mistral-v2-orpo-GGUF 669
- mradermacher/NeuralHyperion-2.0-Mistral-7B-i1-GGUF 574
- mradermacher/CapyLake-7B-v2-laser-i1-GGUF 422
How to load it
from datasets import load_dataset
ds = load_dataset("argilla/distilabel-capybara-dpo-7k-binarized")