---
dataset_info:
  features:
  - name: question
    dtype: string
  - name: answer
    dtype: string
  - name: docid
    dtype: string
  - name: retrieved_docids
    sequence: string
  - name: retrieved_doc_scores
    sequence: float64
  - name: reranked_docids
    sequence: string
  - name: reranked_scores
    sequence: float64
  splits:
  - name: train
    num_bytes: 3084316
    num_examples: 1000
  download_size: 2588779
  dataset_size: 3084316
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
---

This is a subset designed for evaluating Instruction-Aware Contextual Compression[(IACC)](https://github.com/howard-hou/instruction-aware-contextual-compressor/tree/main) performance at various retention ratios.