Dataset · nectar
Nectar
Builder: Berkeley NEST (Starling team)
Availability: available · checked 2026-09-24 · source
Raw record: /data/datasets/nectar.json
Fields
- id
- nectar
- identifiers
- huggingface
- berkeley-nest/Nectar
- builder
- Berkeley NEST (Starling team)
- release_date
- 2023-10-06partial · sourcenote: HF dataset repo creation date, not the builder's own release announcement.
- availability
- available · checked 2026-09-24 · source
- content
- Card: 'the first high-quality 7-wise comparison dataset, generated through GPT-4-based ranking', with prompts from lmsys-chat-1M, ShareGPT, Anthropic/hh-rlhf, UltraFeedback and others; each prompt has ranked answers from several models.recorded · sourcenote: Card license: Apache-2.0 'under the condition that the dataset is not used to compete with OpenAI'.
- primary_sources
- record_history
- date:2026-09-24 · change:created from primary sources (tranche 2 dataset records) · by:wilson-pruitt + claude ·
Parents
Influence without weights
- feedback_from → GPT-4 (Mar. 2023) declared source Card: 'generated through GPT-4-based ranking'. The candidate answers were written by several models; only GPT-4's judgments are asserted here.
Children
Training data
- ← trained_on Starling-RM-7B-alpha declared source Card: 'We train the reward model with preference dataset berkeley-nest/Nectar'. Blog: 'trained with our K-wise loss on the Nectar dataset'.
Read in
No station on the reading path has touched this record yet.