Nectar
A 183K prompt dataset from UC Berkeley with 7 diverse responses per prompt ranked by GPT-4, covering ShareGPT, Alpaca, Open Assistant, LMSYS-Chat, and more. Enables high-quality reward model training with diverse, real-world instruction coverage.
from datasets import load_dataset
ds = load_dataset("berkeley-nest/Nectar")Preview a sample row
{
"prompt": "Describe the key differences between supervised and unsupervised learning.",
"answers": [
{"rank": 1, "answer": "Supervised learning uses labeled data where the model learns...", "source": "gpt-4"},
{"rank": 2, "answer": "In supervised learning, you provide examples with labels...", "source": "claude-1"}
]
}Fine-tune with this dataset
Estimated VRAM to fine-tune with QLoRA (4-bit base model + LoRA adapters), using conservative defaults:
New to fine-tuning? Follow the step-by-step walkthrough: Fine-Tune Your First LLM in 1 Hour
Frequently asked questions
Can I use Nectar commercially?
Yes — Nectar is released under Apache 2.0, a permissive license that allows commercial use, including training models you ship in a product. Check the dataset card for attribution requirements before release.
How much data does Nectar contain, and do I need all of it?
Nectar contains 183K Prompts. You rarely need all of it: for style and format fine-tuning, a few hundred to a few thousand examples are enough — load a slice (e.g. split="train[:1000]") and scale up only if quality plateaus.
What is Nectar best used for?
Reward-model training with 7-way ranked responses. It belongs to the Preference (RLHF / DPO) section of our dataset hub, where you'll find alternatives and complementary sets.