README.md

metadata

pretty_name: Code Space Complexity Tasks
configs:
  - config_name: code_from_sc_mcq
    data_files:
      - split: train
        path: data/code_from_sc_mcq-*.parquet
  - config_name: code_to_sc_mcq
    data_files:
      - split: train
        path: data/code_to_sc_mcq-*.parquet
  - config_name: sc_direct
    data_files:
      - split: train
        path: data/sc_direct-*.parquet
  - config_name: default
    data_files:
      - split: train
        path: data/*.parquet
size_categories:
  - 100K<n<1M
dataset_info:
  features:
    - name: data_source
      dtype: string
    - name: prompt
      list:
        - name: content
          dtype: string
        - name: role
          dtype: string
    - name: ability
      dtype: string
    - name: reward_model
      struct:
        - name: extraction_method
          dtype: string
        - name: ground_truth
          dtype: string
        - name: style
          dtype: string
    - name: extra_info
      struct:
        - name: comp
          dtype: string
        - name: idx
          dtype: string
        - name: problem_id
          dtype: string
        - name: task_type
          dtype: string
  download_size: 482841586
  dataset_size: 493521479

Code Space Complexity prediction

Sourced from BigOBench, using a subset of the N-most commonly occuring space complexities in the dataset

Three tasks:

• sc_direct: Directly predict the space complexity as O(X) given the code
• code_to_sc_mcq: Select the correct space complexity from four options, given the code
• code_from_sc_mcq: Given a space complexity, select the code with that space complexity from four options.

Complexity
O(1)
O(logn)
O(logn*logm)
O(n)
O(n+m)
O(nlogn)
O(n*m)
O(n**2)