Nectar
Fresh
Berkeley NEST's seven-way ranked preference dataset built from GPT-4 rankings over responses from a diverse model pool, used to train Starling.
- Type
- Preference
- Publisher
- Berkeley NEST
- Capabilities
- SafetyInstruction FollowingMulti Turn Dialog
- Runtime
hf_parquet- License
- Apache-2.0
- Size
- 183k prompts, 3.8M response rankings
- Published
- May 2026
Cite
Notes
Only stored in your browser.
Lift evidence
3| Eval | Tools known to lift | Source paper |
|---|---|---|
| MT-Bench | Nectar | - |
| AlpacaEval | Nectar | - |
| Arena-Hard | Nectar | - |
Models
Notable models trained on it
Starling-7B-alphaStarling-LM-7B-betaStarling-RM-7B (reward model)