datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
specter2-corpus-papersneuro-specter2-triplets-multi-pool
Jerjes/neuro-specter2-triplets-multi-pool
This dataset contains anchor papers with their top-K most similar (positive) and most dissimilar (negative) papers based on SPECTER2 embeddings.
Dataset Structure
Each row contains:
anchor_id: Unique identifier for the anchor paper
anchor_title: Title of the anchor paper
anchor_abstract: Abstract of the anchor paper
positive_pool: List of 5 most similar papers, each as [id, title, abstract]
negative_pool: List of 5 most… See the full description on the dataset page: https://huggingface.co/datasets/Jerjes/neuro-specter2-triplets-multi-pool.SPECTER2-data
Dataset Card for "SPECTER2-data"
More Information needed
neuro-specter2-triplets
Jerjes/neuro-specter2-triplets
Triplet dataset for fine-tuning SPECTER2 on neuroscience.
Version date: 2025-08-12
Schema
Columns:
anchor_id, positive_id, negative_id
anchor_title, positive_title, negative_title
anchor_abstract, positive_abstract, negative_abstract
anchor_text, positive_text, negative_text (title + abstract)
Split: train
Load
from datasets import load_dataset
triplets = load_dataset("Jerjes/neuro-specter2-triplets", split="train")
neuro-specter2-triplets-pool
Jerjes/neuro-specter2-triplets-pool
This dataset contains anchor papers with their top-K most similar (positive) and most dissimilar (negative) papers based on SPECTER2 embeddings.
Dataset Structure
Each row contains:
anchor_id: Unique identifier for the anchor paper
anchor_title: Title of the anchor paper
anchor_abstract: Abstract of the anchor paper
positive_pool: List of 5 most similar papers, each as [id, title, abstract]
negative_pool: List of 5 most dissimilar… See the full description on the dataset page: https://huggingface.co/datasets/Jerjes/neuro-specter2-triplets-pool.neuro-specter2-poolsneuro-specter2-sample-data
Jerjes/neuro-specter2-sample-data
This dataset contains anchor papers with their top-K most similar (positive) and most dissimilar (negative) papers based on SPECTER2 embeddings.
Dataset Structure
Each row contains:
anchor_id: Unique identifier for the anchor paper
anchor_title: Title of the anchor paper
anchor_abstract: Abstract of the anchor paper
positive_pool: List of 5 most similar papers, each as [id, title, abstract]
negative_pool: List of 5 most dissimilar… See the full description on the dataset page: https://huggingface.co/datasets/Jerjes/neuro-specter2-sample-data.tldr_vs_abstract_allenai_specter2_aug2023refresh_basetldr_vs_abstract_allenai_specter2_base