UltraFeedback — LLM Preference (RLHF / DPO) Dataset
Large-scale, fine-grained preference dataset. It contains 64k prompts with multiple model responses rated by GPT-4. Essential for RLHF and DPO.
Dataset Details
| Provider | OpenBMB |
| Category | Preference (RLHF / DPO) |
| Size | 64k Rows |
| License | MIT |
| Downloads | 1.2M |
| Tags | RLHF, DPO, Alignment |
from datasets import load_dataset
ds = load_dataset("openbmb/UltraFeedback")
Fine-tune with this dataset
Estimated VRAM to fine-tune with QLoRA (4-bit base model + LoRA adapters), using conservative defaults:
| 7B QLoRA | ~6GB VRAM |
| 13B QLoRA | ~10GB VRAM |
Check if your GPU can fine-tune this →
New to fine-tuning? Follow the step-by-step walkthrough: Fine-Tune Your First LLM in 1 Hour
Related datasets
- OpenHermes 2.5 — The default general-purpose SFT mix for 7B-13B fine-tunes
- HelpSteer2 — Training reward models with fine-grained quality ratings
- Anthropic HH-RLHF — Safety-focused preference training (helpfulness and harmlessness)
- DPO Mix 7K — A small, balanced DPO starter set
Frequently asked questions
Can I use UltraFeedback commercially?
Yes — UltraFeedback is released under MIT, 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 UltraFeedback contain, and do I need all of it?
UltraFeedback contains 64k Rows. 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 UltraFeedback best used for?
The default DPO preference set to run after any SFT pass. It belongs to the Preference (RLHF / DPO) section of our dataset hub, where you'll find alternatives and complementary sets.
← All datasets | Fine-Tuning Guide