Dataset · argilla
Mathematical DPO preferences
argilla/distilabel-math-preference-dpo
English math preference dataset in DPO format; card is nearly empty.
What's inside
A text-generation dataset in English with tabular and text modalities, sized at 1k to 10k examples. The card only contains a "More Information needed" placeholder, so composition, source, and format are not described. Judging by the name, it is a preference-DPO set for math answers.
What you can do with it
- DPO fine-tuning of mathematical LLMs
- Comparing pairs of math solutions
- Checking preferences in answers
- Tasks
- Text generation
- Languages
- English
- Size
- 1K–10K rows
- Downloads this month
- 294
- Likes
- 88
Models trained on it
- TheBloke/SauerkrautLM-SOLAR-Instruct-GGUFText generation 1.4K
- TheBloke/LUNA-SOLARkrautLM-Instruct-GGUFText generation 1.4K
- TheBloke/SauerkrautLM-Mixtral-8x7B-Instruct-GGUFText generation 1.2K
- mradermacher/SauerkrautLM-Mixtral-8x7B-i1-GGUF 955
- mradermacher/SauerkrautLM-Mixtral-8x7B-Instruct-i1-GGUF 788
- h2oai/h2o-danube-1.8b-chatText generation 701
- mradermacher/Sakura-SOLRCA-Math-Instruct-DPO-v1-i1-GGUF 581
- mradermacher/jaskier-7b-dpo-v5.6-i1-GGUF 475
How to load it
from datasets import load_dataset
ds = load_dataset("argilla/distilabel-math-preference-dpo")