---
task_categories:
- text-retrieval
task_ids:
- document-retrieval
config_names:
- corpus
tags:
- text-retrieval
dataset_info:
  - config_name: default
    features:
      - name: query-id
        dtype: string
      - name: corpus-id
        dtype: string
      - name: score
        dtype: float64
  - config_name: corpus
    features:
      - name: id
        dtype: string
      - name: text
        dtype: string
  - config_name: queries
    features:
      - name: id
        dtype: string
      - name: text
        dtype: string
configs:
  - config_name: default
    data_files:
      - split: test
        path: relevance.jsonl
  - config_name: corpus
    data_files:
      - split: corpus
        path: corpus.jsonl
  - config_name: queries
    data_files:
      - split: queries
        path: queries.jsonl
---

DS-1000 is a code generation benchmark with a thousand data science problems spanning seven Python libraries, such as NumPy and Pandas. It employs multi-criteria evaluation metrics, including functional correctness and surface-form constraints, resulting in a high-quality dataset with only 1.8% incorrect solutions among accepted Codex-002 predictions.

**Usage**
```
import datasets

# Download the dataset
queries = datasets.load_dataset("embedding-benchmark/DS1000", "queries")
documents = datasets.load_dataset("embedding-benchmark/DS1000", "corpus")
pair_labels = datasets.load_dataset("embedding-benchmark/DS1000", "default")
```