logic-oriented
Logic-ORiented-Retrieve
Logic-ORiented Retriever Enhancement Dataset
Dataset Description
This dataset is designed for training and evaluating Logic-ORiented Retriever Enhancement (LORE) models.
The dataset implements a three-tier contrastive learning framework with fine-grained sample classification:
P (Positive, label=1): Chunks sufficient to answer the query
N1 (Distractor, label=-1): Chunks used by LLM in query rewriting, seemingly relevant but unhelpful
N2 (Negative, label=0): Other… See the full description on the dataset page: https://huggingface.co/datasets/XiaSheng/Logic-ORiented-Retrieve.Logic-ORiented-Test
XiaSheng/Logic-ORiented-Test
Logic-ORiented Test Dataset - Modified Test
This dataset contains modified test data for three different tasks:
hotpotqa_modified_test: Modified HotpotQA test questions
msmarco_modified_test: Modified MS MARCO test questions
musique_modified_test: Modified MuSiQue test questions
Each split contains questions that have been modified to test logic-oriented retrieval capabilities.
Dataset Structure
The dataset has three splits:… See the full description on the dataset page: https://huggingface.co/datasets/XiaSheng/Logic-ORiented-Test.