Dataset · data-is-better-together
10k prompts with quality ratings
data-is-better-together/10k_prompts_ranked
10,331 community-rated prompts for training and evaluating prompt ranking.
What's inside
The dataset contains 10,331 prompt examples, both synthetic and human-generated, collected from various prompt sets. Quality ratings were provided by 314 open-source ML community participants using Argilla. The English-language data is in tabular and text formats, suited for prompt ranking tasks.
What you can do with it
- Train a reward model or prompt ranking model
- Evaluate an LLM judge on prompt quality assessment tasks
- Use in an RLHF pipeline for prompt selection
- Tasks
- Text classification, Text generation, Reinforcement learning
- Languages
- English
- Size
- 10K–100K rows
- Downloads this month
- 2.4K
- Likes
- 179
How to load it
from datasets import load_dataset
ds = load_dataset("data-is-better-together/10k_prompts_ranked")