README.md

metadata

license: cc-by-nc-sa-4.0
language:
  - en
annotations_creators:
  - no-annotation
task_categories:
  - text-generation
task_ids:
  - language-modeling
size_categories:
  - 10K<n<100K
configs:
  - config_name: python
    data_files:
      - split: test
        path:
          - data/python.jsonl
  - config_name: cc
    data_files:
      - split: test
        path:
          - data/cc.jsonl
  - config_name: arxiv_math
    data_files:
      - split: test
        path:
          - data/arxiv_math.jsonl

This is the compression corpora dataset used in the paper "Compression Represents Intelligence Linearly". We find that LLMs’ intelligence – reflected by benchmark scores – almost linearly correlates with their ability to compress external text corpora. We measure intelligence along three key abilities: knowledge and commonsense, coding, and mathematical reasoning, and provide the corresponding compression corpora here respectively named cc, python, and arxiv_math.

Load the data

from datasets import load_dataset
dataset=load_dataset(r"hkust-nlp/llm-compression",name="python")

print(dataset['test'][0])

More details on compression evaluation are at our github page.

Citation

@misc{huang2024compression,
      title={Compression Represents Intelligence Linearly}, 
      author={Yuzhen Huang and Jinghan Zhang and Zifei Shan and Junxian He},
      year={2024},
      eprint={2404.09937},
      archivePrefix={arXiv},
      primaryClass={cs.CL}
}